Get live statistics and analysis of giyu_codes's profile on X / Twitter

My boss has a network of business owners (150-500 employee size), each with their own teams of AI engineers. He sends me to go audit their AI engineers, as a friendly favor. These companies spent >$5m on teams and projects. 2 years of work. Zero results. 80+ fired & counting.

How I got a job in tech in 30 days. Zero experience -> First job. The purpose of this article is to give you the real ACTIONABLE steps to take on getting a job while providing examples on how I did it. Here are the topics we will cover. The Nuance. The Resume. Your

Recruiter reached out for a position at a startup. Former VPs from Microsoft, EY, and some PE firm.. big investment into them already. Me - my lowest is $250k base, then whatever equity we agree on Recruiter - that's ridiculous and too high, sorry. That's reserved for seniors with 8+ YOE Me - ok np take care ---months go by Recruiter - hey have your changed your mind and lowered your asking salary? It was a bit high first time but just wanted to confirm that is what you meant Me - Still the same. I'm operating at that amount right now. Recruiter- I seriously find that hard to believe.. but ok, I will see what I can do Me - ok.. just confirming that $250k before we start any interviews - 2 weeks later Recruiter- nice, i talked to them and they said they can run three interviews Me - nice, at $250k base? Recruiter - silent CTO interview goes well, says they can meet me at $250k Interview with devs goes well, says they would love to have me onboard. Three days later... Recruiter - sorry, best I could get you is $180k. They don't negotiate. But we have the best benefits package you can find in all of Silicon Valley No office stipend, no sign on bonus, bring your own laptop. WTF just happened?

Turns out you can just ask for things. Called up a client and asked if he was comfortable with me raising the rates from $40k a month to $60k. he said no but at least I asked yeah that's it nothing else lmao but it was worth a shot

The sad truth. If you want to earn the salary of the top 1% then you need to work harder than those in the top 1% so you can catch up to them. Highly recommend upscaling yourself with a very niche skill or a broad and effective set of skills so you can be undeniably great at what you do. You do not need to have a strong network right now. You do not need to have the best tools that cost $200 a month to start coding. You do not need $1,000 in investments to start your side project. You just need to start somewhere. The tl;dr of how I did it was hopping on Gemini ( ai dot dev) and just keeping my head down in work for 2 years. Work for free, work for pennies, work for proof of skill, and post about it online. Then maybe, if you're lucky, you'll get a nice paying job. I went from $0 a month to $400, then $2000, hit $10k for two months then was let go and back to $3000. Finally I got a job and I am at roughly $250k salary. The grind is worth it if you persevere. I wish I could make it so that you didn't have to suffer for two years like I did though.. if I could find a way to condense all the nuance and information that I learned into 3 months and share it, I would... But there is no other advice than for you to just get started, because someone else out there is already working 10x harder for your future job.

Looking back a year ago, the sacrifices were worth it. My girlfriend will soon be my fiancé, I take breaks on weekends now, and I can finally give my partner what she deserves. I received a job offer for "Head of AI" at a law firm. $350k base salary and bonuses tied directly to the value I would bring. I stopped saying "I graduated in 2023" Now I say, "Here is what I have provided in value to these companies." Whether or not I think I am ready for this position, it would be better to try it and grow into it than to just let myself stay complacent.

is everyone lying? like I am flabbergasted at the lessons I am learning. convos from southern california SMEs and who they work with - get a job? white lies on resume. - $1,000,000 ARR? you can say that if you make $85k in one month ( but who cares about the following months, right?) - "we build full AI automation from end to end for large enterprises in southern california" three jupyter notebooks with nothing being automated - "We have 10 customers who are making $5,000,000 in revenue just WAITING for our product" two customers said they like my idea but don't want to invest - "We are pending $2,000,000 in investments" we applied for a loan WHAT THE HECK IS HAPPENING IN BUSINESS MAAAAAAN

Salary ranges so far from the jobs I have interviews for $120k, $160k, $240k It's fucking lit man. Zero lying. Just playing the same game now.

Startups at SF: - 996 - $180k base salary with equity that never pays out - "founding engineer" with no business impact SMEs in OC: - $250k base + bonuses based on value given - impact executive decision making - build many cool apps - true unlimited PTO Why pick startups?

How I Upscaled myself with AI: A Journey from Graduating in 2023 to $100,000+ earned as a Full Stack AI/ML developer & consultant. Long post, but I will teach you how to UPSCALE yourself up with AI. If you don't know me, feel free to read the pinned article in my profile about how I got a job in 30 days. Spoiler alert: I graduated in 2023 and got a “job offer” in Fall 2024 for a job that would start in Spring 2025. I quit that job before I even started and instead continued full-time consulting/software building and doing contract work for clients. But, I am still trying to get into the industry, but only for the right opportunities (Google, please I love Gemini). Here's what I will cover in this post: ---- Prelude ---- Upscaling with AI: Staying Fundamentally Driven ---- Complacency is your downfall. ---- My $100,000 Results ---- Prelude: It's Fall 2022, my senior year of University. Layoffs are rising, ChatGPT is released, and I have a terrible portfolio with no real projects. I built a foundation from four years of schooling, but so what? C, C++, Java, Assembly, Python, HTML/CSS/JavaScript, R, Differential Equations, Computer Networking, Calculus, Discrete Math... you get the point. My passion was in Data Science & AI/ML, working on projects that had you label data, run analysis, and train a model on my cute little 3060ti at home. But what does that matter if ChatGPT just came out and absolutely STOMPED my future into the ground? (or so I thought…) Late December 2022, around my 28th birthday, my girlfriend predicted the market for me. She said this would happen: - More layoffs. - Tons of competition (from laid-off workers and recent grads). - AI is going to take your job if you don’t do anything about it. So, Spring 2023. I'm heading into my final semester with: - Zero internships. - Zero job offers. - Zero marketable skills. Layoffs hit an all time peak in Spring 2023, something like 250k laid off. I basically resigned to my fate as a McDonalds worker, just putting the fries in the bag.. (No disrespect to fast food workers – it's honest work – but after four years of a CS degree? Come on man I’m not going back. Stay humble I guess.) Then, my girlfriend drops another drop of wisdom: "Stop looking at LinkedIn and Indeed. That is where everyone else is looking. Check Craigslist. Do something different." BOOM. I found a contract gig with one of the biggest IT firms in my area. During the interview they said they wanted to build an enterprise-level AI application for their company and for their 150 clients – law firms, schools, real estate, dentist office, you name it. Think about that for a second. A NEW GRAD, with a portfolio full of Jupyter Notebooks and PyGUI scripts, tasked with building a Full-stack AI/ML application for hundreds of businesses. DOOOOMED. Graduation day. I walk off stage, diploma in hand, and my phone rings. It's the CEO. "You're hired. Let's start tomorrow." Literally walking off the field I got that first client. BLESSED. I thought there was absolutely no chance I would be able to pull this off. I was still trying to figure out the difference between NextJS and Vite (WHICH ONE SCAAALES BETTER?!). I was still trying to figure out what truly makes someone a Senior or Principal developer. I spent 4 months figuring out a database schema lmao. No worries though. I just had to lock in and do what I do best - work. I knew the ONLY way I can stay relevant in the midst of all this competition is to get better. I need to UPSCALE myself, QUICK. ---- Upscaling with AI: Staying Fundamentally Driven It wasn't about memorizing some textbook, watching another YouTube tutorial, or chasing the latest shiny tool. This was a mindset. A philosophy. A LIFESTYLE that I needed to understand. I realized that AI is a powerful tool, but it's not a replacement for understanding the fundamentals. So, as of today I spent the last 22 months speed running my career journey by burning at 10-16 hours a day 6-7 days a week THROUGH HOLIDAYS. This is not the level of sacrifice you should go through and I hope this helps prevent that for you because I would not wish it upon anyone. Also, here's the thing: Most people will screw this up. They'll assume AI has the perfect code, that it's fundamentally right. Newsflash: AI can't handle super complex systems. You will run into situations where it fails. So, how do you not become automated by AI? By understanding the WHY behind the code, not just the how. And a lot of that "why" is stuff you could have learned in school, if you weren't just coasting. I will be using the “Enterprise on AI” project from early 2023 as my leading example. I am not going to give you some cookie cutter, "5 steps to success" BS. This is just how I figured my shit out. Frontend: Building an AI chatbot in early 2023 wasn't hard, but it wasn't as accessible as it is now... There weren't many examples or guides that truly showed how to build it from start to finish. How do you fetch messages? How do you send them? How do you connect to a backend? How do you display the messages in order? What should the data model look like? I sorted the conversation messages three times at some point: backend while fetching, backend before sending, and frontend when receiving. OVERKILL AND BLOATED. Most of this stuff wasn’t even specific to AI chatbots too, which is even more embarrassing for me. But my brain thought to only look for AI chatbot examples. I used React, create-react-app to be specific (RIP). I read a bit of the docs and thought I understood React well enough to just start building - this is a good instinct to be honest. Better than tutorial hell. The PROBLEM, was that I asked ChatGPT for help, got some code, and quickly found myself three hours deep into debugging the same thing over and over and over. Messages sent twice, messages not sent on "Enter", and chat freezes after 30 messages. All I'm typing to ChatGPT is: "this code is broken, can you fix it?" -> "THE CODE IS BROKEN, PLEASE FIX" -> "I SWEAR IF YOU DON'T FIX THIS I WILL END -"... yeah you've probably been there too. I took a step back and saw simple things like useEffect being placed too deep in a component that would re-render often. I didn't memoize certain components that could have been memoized. Or to be honest maybe I THOUGHT I understood the component lifecycle but I just didn't? I was letting AI handle the how without understanding the why of React's rendering. Then I spent another 30 minutes re-reading the documentation, looked online for a bit more examples, and typed out the changes myself after asking ChatGPT for examples of re-rendering and what to look out for in my react code. The questions changed from "Why is this broken?" to "Explain the difference between using useEffect to fetch data and why I shouldn’t over use it?", or even “What is the component life cycle and how does the DOM really work?” and even then I had more to learn - But I broke the cycle of endless debugging and trusting AI too much when I could have just read it myself and trusted myself to understand the BASICS. My root issue was not going to be cured by generating more code. it was going to be cured by taking the time to ask questions and UNDERSTANDING the topics with the help of AI. After it was all over, I made sure to take notes and save it so I can learn from my mistakes. This is something you should do because it can help future teammates from wasting time on the same issue ( WINK WINK THIS IS WHAT MAKES YOU A GOOD TEAMMATE IN THE INDUSTRY, at least that is what I have been told but I can't confirm as I have not worked in it yet) Backend: I was using Neo4j because I was familiar with it during my University days and graphs made sense for highly connected SME data. Vector embeddings were supported by Neo4j and it was becoming an increasingly requested feature to improve on across all database ecosystems. Eventually I switched to MongoDB and Qdrant because they were new shiny tools. But it didn't matter, I still ended up fighting against AI in understanding and implementing how these databases should fundamentally work. I was trying to store 500 emails at once in Neo4j - this is a small task for data ingestion, right? To some this is basic and pathetically simple, but you have to understand that the ones who think this are the ones might have forgotten what its like to learn this stuff and gain the intuition. But yeah I struggled with this early on lol. Transactions are a thing and if the creation of a node failed then they all failed. I didn't really understand ACID to the level of care that I needed to, but then I saw that I could add a keyword argument that basically said "if one fails still push up the rest". Still not the best, but it beats my first few tries of just “for looping” individual node creations. I knew about transactions and how databases worked, but I hadn't connected the dots to the practical implications of data integrity. Thankfully I gave a shit about my database class, otherwise I wouldn’t have even known what types of questions to ask. But as I kept asking for fundamental questions, I found myself stuck in loops of debugging outdated syntax and fighting against AI that did not have the latest information. ChatGPT did not have much support for Neo4j, MongoDB, or Qdrant as they were all rolling out new features. I needed to learn how to optimize queries because data ingestion was taking like 5-10 minutes each time a user would connect their data. I looked up community examples and documentation to learn how to use "CALL" in Neo4j or how to properly batch up documents in MongoDB, or even simply using list comprehensions and asynchronous libraries in python - they all made the difference in the performance of the application. I had to move beyond AI-generated code and understand why certain approaches were faster and more efficient. As I came across each topic I would ask: "What is the difference between concurrency vs parallelism in this case?" -> "What are the low level costs of indexing a property? Is it bad to have too many indexes?" All of these topics are covered in school, so understanding them helped. But you are removing the opportunity of gaining INTUITION and connecting your two brain cells if you just let ChatGPT do the work for you. Cloud shenanigans: Cloud networking was overwhelming when I got into it. I'm not even going to pretend I'm anywhere close to an expert. This is where the difference between "knowing the terms" and "actually making it work without setting everything on fire" becomes painfully clear. I Dockerized my backend, pushed it up to ECS, and then launched an EC2 instance behind a VPC. That one sentence? Weeks. WEEKS of figuring out the little nuances that NOBODY tells you about, but you just have to do it. Fighting with the AWS console, setting up IAM permissions, and using the terminal… let alone just staring at all that documentation. It was overwhelming, but I needed to do it. ChatGPT was useless for the account setup – the stuff that isn't code, but about understanding how each Cloud provider works, getting used to the overwhelming feeling of the web page, and relentlessly digging for information. Asking AI "What's a VPC?" gets you a definition, not a solution. I knew about security, theoretically. But I hadn't fully connected it to the reality of VPCs, security groups, and all that. Result? I almost left my servers wide open. Ports, protocols, network topology – all are familiar in my brain because of my Computer Networking class but you need to go through the motions to get the intuition. There was a point in time where my Blizzard app was downloading Call of Duty and I didn’t realize it was blocking Neo4j. Why? Same port, apparently. But Wireshark saved my ass because I learned how to inspect packets in school. That was a school fundamental saving me. A direct application of something I'd learned from school, WOOOW CRAZY. Lesson? Docs first. Fundamentals FIRST. Then AI. "Why care about network security if my app is secure?" becomes a different question after you (sort of) understand VPCs. Also, even understanding latency was a simple thing in hindsight, but an easy thing to miss. Frontend hosted in California? Backend hosted in New York? 90ms delay. So you think the time to first token is long but really it was your infrastructure. Also friendly reminder to clean up your logs other CloudWatch will laugh at you, smack you, and take your money. AI/ML: Applied LLMs for Full Stack apps in 2022 was basically a free for all. I knew about embeddings, semantic search, KNNs, not training on test set, how to make a neural network from scratch, tranformers, and a ton of AI/ML stuff at a pretty granular level. It was my passion in school, so I was glad I had a decent advantage. Vector embeddings, hybrid search, chunking, AI as a Judge, rerankers - most people just used these chucked these buzzwords everywhere and there were not many real world examples to help you learn what it is like to put these into an application. So you just had to do it yourself. I am CONVINCED that anyone who was doing this around this time laughed when “AI Agents” came out because we were already making agents by early 2023. But even then my lack of experience hit like a brick wall. How do you retrieve data for different users, with the same question? How do you protect sensitive data?! YES IT WAS THAT BAD. But I had to learn. Also, for what its worth I told the client this and he was patient with it. Truly the GOAT. Anyways, I was using tools without understanding the nuances. I knew the theory, but I hadn't dealt with the practical challenges of relevance and context. YouTube, skimming blogs. Not enough. I was living and breathing Jason Liu’s early RAG videos and office hours and even then I didn’t have the intuition to really build out a full information and recommendation retrieval system. I started asking better questions: "Why is just vector similarity often garbage?" -> "How do I combine search methods without the database supporting it?" -> "How do I know if the results are good?" -> “What is a proper evaluation method when you get 100 data results and each user expects different data points?” I needed to understand not just how to use embeddings, but when and why they worked (or didn't). I had to move beyond "magic black box" to understanding complex information retrieval systems. By reading docs, research papers, and joining communities, I was able to ask better questions and quickly found myself finding a rhythm. Not just using AI for outputs, but using my intuition to find better questions to ask AI. If you find yourself relying on AI to generate the right answers for you instead of generating answers you already know, you are going to have a bad time. The Flow State: It wasn't linear. All of this happened at once: learning, failing, succeeding. But a pattern did show up! 1. Screw Up: Run into a wall, usually because I missed a fundamental or didn't apply what I should have known. 2. Read the Docs (Finally): Actually read the documentation. 3. Ask Smarter Questions: Use that to ask AI questions that weren't useless. 4. Learn & Iterate: Use AI's answers, fix it, write it down. 5. Repeat. Constantly. Also I tricked you lol this is totally 5 steps and they are purposefully generic because if you didn’t read the content you won’t understand the intuition of how you get there. Have you ever noticed how a distillation of information always results in hindsight light bulbs? Yeah that’s you automating yourself away. The reason I say “Upscale” and not “Upskill” is because you can learn so much if you just ask the right questions and stop being complacent. School is useful – not just for the paper, but for the fundamentals – if you apply it. If you didn’t go to school, you’re still going to make it. Because you don’t need a degree to give a shit. Don't coast. Don't let AI think for you. Use it to think better and get through problems quicker. ---- Complacency is your downfall. This isn't just for new grads. This is for anyone. Devs, business owners, literally ANYONE – who thinks they can coast. I've seen it. Developer friends at small companies, too comfortable, not learning. They get laid off, no side projects, screwed. They didn't use AI to UPSCALE themselves; they let it make them obsolete. Complacency kills careers, regardless of AI existing or not. I've seen businesses lose clients because another company offered better solutions, PoWereD By AI. Or worse, they didn't bother to ask simple questions. You don't need a whole new app. Ask ChatGPT: "How can I improve my business? Here's what I do… {business info here}" Be proactive, not reactive. Hot take: The people working in the industry right now, the ones who have not had the time to catch the AI wave, THEY ARE THE SCARIEST ONES. NONE OF THIS IS NEW TO THEM (at least the good ones). Once they have time to catch up, it’s over for me. So I need to keep UPSCALING myself to stay relevant. Strong fundamentals + AI = super-powered generalist, fast. Complacent? Just put the fries in the bag (again not a bad thing, I’ll probably be there soon tbh) ---- My $100,000+ Results (Not a Brag, a Warning) I am grateful and humbled to say I barely survived and made a living over the last 22 months. From back then to January 1st 2025 I made over $100,000 (taxes are going to ruin me). Compared to then, I am now a way better developer but on top of that I am an even better problem solver. You can just CARE about things. Also, this is not a flex. THIS IS A WARNING. I was not able to get an internship during my time in school and I did not graduate with a job. Just a 28 year old dude (now 30) that was forced to find ways to put food on the table for me, my girlfriend, and our cat. IT SUCKED. IT SUCKED ASS MAN. We sacrificed so much and I wish I had me from the future tell me to do something about it sooner. Somehow, I, a zero-credential chill guy, was able to build a portfolio of successful cases (and failures) in which I impacted local Small-Medium Enterprise businesses. It all started with the Enterprise on AI application where I got to see the landscape of so many different companies and I worked my ass off to understand how to truly impact them. It wasn’t in the shiny code tools, it wasn’t because I learned how to use Server Side Rendering in Nextjs, and it wasn’t because I spent time picking a new Agent framework because it promised to SCALE better. It was because I wasn’t afraid to ask questions, I wasn’t afraid to look stupid, and I put the tech second to finding solutions. Everyone you are competing with is coding up some new SaaS and showcasing it. I get it, ITS OVERWHELMING. HOW DO YOU KEEP UP? You need to find a way to make yourself stand out and push ahead. I did so by removing all complacency that was holding me back and just grinded away. I used my life experience and background to help bridge the gap of my lack of “Experience” in the tech industry. My fundamental background helped me get ahead. I studied over 150 different businesses within over 30 different industries that I didn't even know existed, AND interviewed a few dozen CEOs. But what good is that experience if I don't DO ANYTHING WITH IT? The same way those who went to school think back as if school wasn't valuable. The same way the people in the industry with no degree think that they will be replaced. You just need to find your fundamental value and run with it. Because your background won't matter. Just DO things. Don't get me wrong though I didn't implement specific AI workflows for all of them. I had a chance to research them and understand their needs and did some work that would impact all of them at general levels and tackled a few specific workflows for their businesses... but not all of them lol. At this point I have only helped <20 SMEs directly… It’s interesting knowing that I have directly impacted millions and millions in revenue - whether it was from proposal automation, analyzing phone call sentiment to improve customer relations, or improving Lead Generation workflows to get new clients. It does come with a lot of downsides too though... Watching people get laid off because I “automated” their jobs (stupid choice by businesses in my opinion. You should be ENHANCING individuals, not outright replacing them). I led teams of devs with WAY more experience than me and I'm grateful that they gave me more respect than I deserved. All this while still knowing my place though.. It was tricky to manage. I was humbled and mentored multiple times over... I just wanted to LEARN things. Now? Titles don't really matter at the end of the day. People just want to work with people that can respect the fact that we are all working towards the same goal. I am probably a junior engineer in some cases, a principal in others, or maybe at best an intern? I don’t really know. The real warning: Don’t be complacent. Stop assuming so much of AI and start reviewing the fundamentals as you learn things. I wrote this out of frustration of some peers who accepted defeat within the tech job market and also out of frustration of recent experiences in the business world... I just don't want people wasting time by making the same mistakes I made. I hate watching people stay complacent. You have an open opportunity to UPSCALE yourself but you choose to “Accept” vibe code your way to automation without first learning the fundamentals. The ones who can vibe code their way to success have years of experience under their belt to allow them to do that, whether they know it or not. Tl;dr - eat a chipotle bowl for me. I'm not summarizing this. You will likely run this through AI to summarize this and that is OK with me. As long as you put in the work at some point in time down the line.

shoutout to my girlfriend who has given me new life. in 7 days we will have spent 10 years together (long story, new followers scroll my posts for more details) she's the reason I can wake up at 6am and sleep at 12am ABSOLUTELY GRINDING away so I can scale myself up to getting a job in the tech industry. I'm almost there and she will be absolutely spoiled with every penny I make.

I got reached out to by Meta. 😳 Is it time?

At this point I genuinely tell recruiters my minimum is $350k base salary and they see that I graduated two years ago and just wonder wtf my problem is. My problem is that they're not offering enough.

Hi everyone I can't post articles but I just want to give free info so here you go: Here is what I learned from building an "AI on Small-Medium Enterprise" app as the sole Full Stack AI/ML developer. This helped one client, who manages over 150+ businesses, save significant time and money by leveraging their own operational intelligence. post 1/??

Recruiter reached out for a position at a startup. Former VPs from Microsoft, EY, and some PE firm.. big investment into them already. Me - my lowest is $250k base, then whatever equity we agree on Recruiter - that's ridiculous and too high, sorry. That's reserved for seniors with 8+ YOE Me - ok np take care ---months go by Recruiter - hey have your changed your mind and lowered your asking salary? It was a bit high first time but just wanted to confirm that is what you meant Me - Still the same. I'm operating at that amount right now. Recruiter- I seriously find that hard to believe.. but ok, I will see what I can do Me - ok.. just confirming that $250k before we start any interviews - 2 weeks later Recruiter- nice, i talked to them and they said they can run three interviews Me - nice, at $250k base? Recruiter - silent CTO interview goes well, says they can meet me at $250k Interview with devs goes well, says they would love to have me onboard. Three days later... Recruiter - sorry, best I could get you is $180k. They don't negotiate. But we have the best benefits package you can find in all of Silicon Valley No office stipend, no sign on bonus, bring your own laptop. WTF just happened?

My boss has a network of business owners (150-500 employee size), each with their own teams of AI engineers. He sends me to go audit their AI engineers, as a friendly favor. These companies spent >$5m on teams and projects. 2 years of work. Zero results. 80+ fired & counting.

How I got a job in tech in 30 days. Zero experience -> First job. The purpose of this article is to give you the real ACTIONABLE steps to take on getting a job while providing examples on how I did it. Here are the topics we will cover. The Nuance. The Resume. Your

what do I do for the next 13 days?

Looking back a year ago, the sacrifices were worth it. My girlfriend will soon be my fiancé, I take breaks on weekends now, and I can finally give my partner what she deserves. I received a job offer for "Head of AI" at a law firm. $350k base salary and bonuses tied directly to the value I would bring. I stopped saying "I graduated in 2023" Now I say, "Here is what I have provided in value to these companies." Whether or not I think I am ready for this position, it would be better to try it and grow into it than to just let myself stay complacent.

I got reached out to by Meta. 😳 Is it time?

Turns out you can just ask for things. Called up a client and asked if he was comfortable with me raising the rates from $40k a month to $60k. he said no but at least I asked yeah that's it nothing else lmao but it was worth a shot

is everyone lying? like I am flabbergasted at the lessons I am learning. convos from southern california SMEs and who they work with - get a job? white lies on resume. - $1,000,000 ARR? you can say that if you make $85k in one month ( but who cares about the following months, right?) - "we build full AI automation from end to end for large enterprises in southern california" three jupyter notebooks with nothing being automated - "We have 10 customers who are making $5,000,000 in revenue just WAITING for our product" two customers said they like my idea but don't want to invest - "We are pending $2,000,000 in investments" we applied for a loan WHAT THE HECK IS HAPPENING IN BUSINESS MAAAAAAN

How I Upscaled myself with AI: A Journey from Graduating in 2023 to $100,000+ earned as a Full Stack AI/ML developer & consultant. Long post, but I will teach you how to UPSCALE yourself up with AI. If you don't know me, feel free to read the pinned article in my profile about how I got a job in 30 days. Spoiler alert: I graduated in 2023 and got a “job offer” in Fall 2024 for a job that would start in Spring 2025. I quit that job before I even started and instead continued full-time consulting/software building and doing contract work for clients. But, I am still trying to get into the industry, but only for the right opportunities (Google, please I love Gemini). Here's what I will cover in this post: ---- Prelude ---- Upscaling with AI: Staying Fundamentally Driven ---- Complacency is your downfall. ---- My $100,000 Results ---- Prelude: It's Fall 2022, my senior year of University. Layoffs are rising, ChatGPT is released, and I have a terrible portfolio with no real projects. I built a foundation from four years of schooling, but so what? C, C++, Java, Assembly, Python, HTML/CSS/JavaScript, R, Differential Equations, Computer Networking, Calculus, Discrete Math... you get the point. My passion was in Data Science & AI/ML, working on projects that had you label data, run analysis, and train a model on my cute little 3060ti at home. But what does that matter if ChatGPT just came out and absolutely STOMPED my future into the ground? (or so I thought…) Late December 2022, around my 28th birthday, my girlfriend predicted the market for me. She said this would happen: - More layoffs. - Tons of competition (from laid-off workers and recent grads). - AI is going to take your job if you don’t do anything about it. So, Spring 2023. I'm heading into my final semester with: - Zero internships. - Zero job offers. - Zero marketable skills. Layoffs hit an all time peak in Spring 2023, something like 250k laid off. I basically resigned to my fate as a McDonalds worker, just putting the fries in the bag.. (No disrespect to fast food workers – it's honest work – but after four years of a CS degree? Come on man I’m not going back. Stay humble I guess.) Then, my girlfriend drops another drop of wisdom: "Stop looking at LinkedIn and Indeed. That is where everyone else is looking. Check Craigslist. Do something different." BOOM. I found a contract gig with one of the biggest IT firms in my area. During the interview they said they wanted to build an enterprise-level AI application for their company and for their 150 clients – law firms, schools, real estate, dentist office, you name it. Think about that for a second. A NEW GRAD, with a portfolio full of Jupyter Notebooks and PyGUI scripts, tasked with building a Full-stack AI/ML application for hundreds of businesses. DOOOOMED. Graduation day. I walk off stage, diploma in hand, and my phone rings. It's the CEO. "You're hired. Let's start tomorrow." Literally walking off the field I got that first client. BLESSED. I thought there was absolutely no chance I would be able to pull this off. I was still trying to figure out the difference between NextJS and Vite (WHICH ONE SCAAALES BETTER?!). I was still trying to figure out what truly makes someone a Senior or Principal developer. I spent 4 months figuring out a database schema lmao. No worries though. I just had to lock in and do what I do best - work. I knew the ONLY way I can stay relevant in the midst of all this competition is to get better. I need to UPSCALE myself, QUICK. ---- Upscaling with AI: Staying Fundamentally Driven It wasn't about memorizing some textbook, watching another YouTube tutorial, or chasing the latest shiny tool. This was a mindset. A philosophy. A LIFESTYLE that I needed to understand. I realized that AI is a powerful tool, but it's not a replacement for understanding the fundamentals. So, as of today I spent the last 22 months speed running my career journey by burning at 10-16 hours a day 6-7 days a week THROUGH HOLIDAYS. This is not the level of sacrifice you should go through and I hope this helps prevent that for you because I would not wish it upon anyone. Also, here's the thing: Most people will screw this up. They'll assume AI has the perfect code, that it's fundamentally right. Newsflash: AI can't handle super complex systems. You will run into situations where it fails. So, how do you not become automated by AI? By understanding the WHY behind the code, not just the how. And a lot of that "why" is stuff you could have learned in school, if you weren't just coasting. I will be using the “Enterprise on AI” project from early 2023 as my leading example. I am not going to give you some cookie cutter, "5 steps to success" BS. This is just how I figured my shit out. Frontend: Building an AI chatbot in early 2023 wasn't hard, but it wasn't as accessible as it is now... There weren't many examples or guides that truly showed how to build it from start to finish. How do you fetch messages? How do you send them? How do you connect to a backend? How do you display the messages in order? What should the data model look like? I sorted the conversation messages three times at some point: backend while fetching, backend before sending, and frontend when receiving. OVERKILL AND BLOATED. Most of this stuff wasn’t even specific to AI chatbots too, which is even more embarrassing for me. But my brain thought to only look for AI chatbot examples. I used React, create-react-app to be specific (RIP). I read a bit of the docs and thought I understood React well enough to just start building - this is a good instinct to be honest. Better than tutorial hell. The PROBLEM, was that I asked ChatGPT for help, got some code, and quickly found myself three hours deep into debugging the same thing over and over and over. Messages sent twice, messages not sent on "Enter", and chat freezes after 30 messages. All I'm typing to ChatGPT is: "this code is broken, can you fix it?" -> "THE CODE IS BROKEN, PLEASE FIX" -> "I SWEAR IF YOU DON'T FIX THIS I WILL END -"... yeah you've probably been there too. I took a step back and saw simple things like useEffect being placed too deep in a component that would re-render often. I didn't memoize certain components that could have been memoized. Or to be honest maybe I THOUGHT I understood the component lifecycle but I just didn't? I was letting AI handle the how without understanding the why of React's rendering. Then I spent another 30 minutes re-reading the documentation, looked online for a bit more examples, and typed out the changes myself after asking ChatGPT for examples of re-rendering and what to look out for in my react code. The questions changed from "Why is this broken?" to "Explain the difference between using useEffect to fetch data and why I shouldn’t over use it?", or even “What is the component life cycle and how does the DOM really work?” and even then I had more to learn - But I broke the cycle of endless debugging and trusting AI too much when I could have just read it myself and trusted myself to understand the BASICS. My root issue was not going to be cured by generating more code. it was going to be cured by taking the time to ask questions and UNDERSTANDING the topics with the help of AI. After it was all over, I made sure to take notes and save it so I can learn from my mistakes. This is something you should do because it can help future teammates from wasting time on the same issue ( WINK WINK THIS IS WHAT MAKES YOU A GOOD TEAMMATE IN THE INDUSTRY, at least that is what I have been told but I can't confirm as I have not worked in it yet) Backend: I was using Neo4j because I was familiar with it during my University days and graphs made sense for highly connected SME data. Vector embeddings were supported by Neo4j and it was becoming an increasingly requested feature to improve on across all database ecosystems. Eventually I switched to MongoDB and Qdrant because they were new shiny tools. But it didn't matter, I still ended up fighting against AI in understanding and implementing how these databases should fundamentally work. I was trying to store 500 emails at once in Neo4j - this is a small task for data ingestion, right? To some this is basic and pathetically simple, but you have to understand that the ones who think this are the ones might have forgotten what its like to learn this stuff and gain the intuition. But yeah I struggled with this early on lol. Transactions are a thing and if the creation of a node failed then they all failed. I didn't really understand ACID to the level of care that I needed to, but then I saw that I could add a keyword argument that basically said "if one fails still push up the rest". Still not the best, but it beats my first few tries of just “for looping” individual node creations. I knew about transactions and how databases worked, but I hadn't connected the dots to the practical implications of data integrity. Thankfully I gave a shit about my database class, otherwise I wouldn’t have even known what types of questions to ask. But as I kept asking for fundamental questions, I found myself stuck in loops of debugging outdated syntax and fighting against AI that did not have the latest information. ChatGPT did not have much support for Neo4j, MongoDB, or Qdrant as they were all rolling out new features. I needed to learn how to optimize queries because data ingestion was taking like 5-10 minutes each time a user would connect their data. I looked up community examples and documentation to learn how to use "CALL" in Neo4j or how to properly batch up documents in MongoDB, or even simply using list comprehensions and asynchronous libraries in python - they all made the difference in the performance of the application. I had to move beyond AI-generated code and understand why certain approaches were faster and more efficient. As I came across each topic I would ask: "What is the difference between concurrency vs parallelism in this case?" -> "What are the low level costs of indexing a property? Is it bad to have too many indexes?" All of these topics are covered in school, so understanding them helped. But you are removing the opportunity of gaining INTUITION and connecting your two brain cells if you just let ChatGPT do the work for you. Cloud shenanigans: Cloud networking was overwhelming when I got into it. I'm not even going to pretend I'm anywhere close to an expert. This is where the difference between "knowing the terms" and "actually making it work without setting everything on fire" becomes painfully clear. I Dockerized my backend, pushed it up to ECS, and then launched an EC2 instance behind a VPC. That one sentence? Weeks. WEEKS of figuring out the little nuances that NOBODY tells you about, but you just have to do it. Fighting with the AWS console, setting up IAM permissions, and using the terminal… let alone just staring at all that documentation. It was overwhelming, but I needed to do it. ChatGPT was useless for the account setup – the stuff that isn't code, but about understanding how each Cloud provider works, getting used to the overwhelming feeling of the web page, and relentlessly digging for information. Asking AI "What's a VPC?" gets you a definition, not a solution. I knew about security, theoretically. But I hadn't fully connected it to the reality of VPCs, security groups, and all that. Result? I almost left my servers wide open. Ports, protocols, network topology – all are familiar in my brain because of my Computer Networking class but you need to go through the motions to get the intuition. There was a point in time where my Blizzard app was downloading Call of Duty and I didn’t realize it was blocking Neo4j. Why? Same port, apparently. But Wireshark saved my ass because I learned how to inspect packets in school. That was a school fundamental saving me. A direct application of something I'd learned from school, WOOOW CRAZY. Lesson? Docs first. Fundamentals FIRST. Then AI. "Why care about network security if my app is secure?" becomes a different question after you (sort of) understand VPCs. Also, even understanding latency was a simple thing in hindsight, but an easy thing to miss. Frontend hosted in California? Backend hosted in New York? 90ms delay. So you think the time to first token is long but really it was your infrastructure. Also friendly reminder to clean up your logs other CloudWatch will laugh at you, smack you, and take your money. AI/ML: Applied LLMs for Full Stack apps in 2022 was basically a free for all. I knew about embeddings, semantic search, KNNs, not training on test set, how to make a neural network from scratch, tranformers, and a ton of AI/ML stuff at a pretty granular level. It was my passion in school, so I was glad I had a decent advantage. Vector embeddings, hybrid search, chunking, AI as a Judge, rerankers - most people just used these chucked these buzzwords everywhere and there were not many real world examples to help you learn what it is like to put these into an application. So you just had to do it yourself. I am CONVINCED that anyone who was doing this around this time laughed when “AI Agents” came out because we were already making agents by early 2023. But even then my lack of experience hit like a brick wall. How do you retrieve data for different users, with the same question? How do you protect sensitive data?! YES IT WAS THAT BAD. But I had to learn. Also, for what its worth I told the client this and he was patient with it. Truly the GOAT. Anyways, I was using tools without understanding the nuances. I knew the theory, but I hadn't dealt with the practical challenges of relevance and context. YouTube, skimming blogs. Not enough. I was living and breathing Jason Liu’s early RAG videos and office hours and even then I didn’t have the intuition to really build out a full information and recommendation retrieval system. I started asking better questions: "Why is just vector similarity often garbage?" -> "How do I combine search methods without the database supporting it?" -> "How do I know if the results are good?" -> “What is a proper evaluation method when you get 100 data results and each user expects different data points?” I needed to understand not just how to use embeddings, but when and why they worked (or didn't). I had to move beyond "magic black box" to understanding complex information retrieval systems. By reading docs, research papers, and joining communities, I was able to ask better questions and quickly found myself finding a rhythm. Not just using AI for outputs, but using my intuition to find better questions to ask AI. If you find yourself relying on AI to generate the right answers for you instead of generating answers you already know, you are going to have a bad time. The Flow State: It wasn't linear. All of this happened at once: learning, failing, succeeding. But a pattern did show up! 1. Screw Up: Run into a wall, usually because I missed a fundamental or didn't apply what I should have known. 2. Read the Docs (Finally): Actually read the documentation. 3. Ask Smarter Questions: Use that to ask AI questions that weren't useless. 4. Learn & Iterate: Use AI's answers, fix it, write it down. 5. Repeat. Constantly. Also I tricked you lol this is totally 5 steps and they are purposefully generic because if you didn’t read the content you won’t understand the intuition of how you get there. Have you ever noticed how a distillation of information always results in hindsight light bulbs? Yeah that’s you automating yourself away. The reason I say “Upscale” and not “Upskill” is because you can learn so much if you just ask the right questions and stop being complacent. School is useful – not just for the paper, but for the fundamentals – if you apply it. If you didn’t go to school, you’re still going to make it. Because you don’t need a degree to give a shit. Don't coast. Don't let AI think for you. Use it to think better and get through problems quicker. ---- Complacency is your downfall. This isn't just for new grads. This is for anyone. Devs, business owners, literally ANYONE – who thinks they can coast. I've seen it. Developer friends at small companies, too comfortable, not learning. They get laid off, no side projects, screwed. They didn't use AI to UPSCALE themselves; they let it make them obsolete. Complacency kills careers, regardless of AI existing or not. I've seen businesses lose clients because another company offered better solutions, PoWereD By AI. Or worse, they didn't bother to ask simple questions. You don't need a whole new app. Ask ChatGPT: "How can I improve my business? Here's what I do… {business info here}" Be proactive, not reactive. Hot take: The people working in the industry right now, the ones who have not had the time to catch the AI wave, THEY ARE THE SCARIEST ONES. NONE OF THIS IS NEW TO THEM (at least the good ones). Once they have time to catch up, it’s over for me. So I need to keep UPSCALING myself to stay relevant. Strong fundamentals + AI = super-powered generalist, fast. Complacent? Just put the fries in the bag (again not a bad thing, I’ll probably be there soon tbh) ---- My $100,000+ Results (Not a Brag, a Warning) I am grateful and humbled to say I barely survived and made a living over the last 22 months. From back then to January 1st 2025 I made over $100,000 (taxes are going to ruin me). Compared to then, I am now a way better developer but on top of that I am an even better problem solver. You can just CARE about things. Also, this is not a flex. THIS IS A WARNING. I was not able to get an internship during my time in school and I did not graduate with a job. Just a 28 year old dude (now 30) that was forced to find ways to put food on the table for me, my girlfriend, and our cat. IT SUCKED. IT SUCKED ASS MAN. We sacrificed so much and I wish I had me from the future tell me to do something about it sooner. Somehow, I, a zero-credential chill guy, was able to build a portfolio of successful cases (and failures) in which I impacted local Small-Medium Enterprise businesses. It all started with the Enterprise on AI application where I got to see the landscape of so many different companies and I worked my ass off to understand how to truly impact them. It wasn’t in the shiny code tools, it wasn’t because I learned how to use Server Side Rendering in Nextjs, and it wasn’t because I spent time picking a new Agent framework because it promised to SCALE better. It was because I wasn’t afraid to ask questions, I wasn’t afraid to look stupid, and I put the tech second to finding solutions. Everyone you are competing with is coding up some new SaaS and showcasing it. I get it, ITS OVERWHELMING. HOW DO YOU KEEP UP? You need to find a way to make yourself stand out and push ahead. I did so by removing all complacency that was holding me back and just grinded away. I used my life experience and background to help bridge the gap of my lack of “Experience” in the tech industry. My fundamental background helped me get ahead. I studied over 150 different businesses within over 30 different industries that I didn't even know existed, AND interviewed a few dozen CEOs. But what good is that experience if I don't DO ANYTHING WITH IT? The same way those who went to school think back as if school wasn't valuable. The same way the people in the industry with no degree think that they will be replaced. You just need to find your fundamental value and run with it. Because your background won't matter. Just DO things. Don't get me wrong though I didn't implement specific AI workflows for all of them. I had a chance to research them and understand their needs and did some work that would impact all of them at general levels and tackled a few specific workflows for their businesses... but not all of them lol. At this point I have only helped <20 SMEs directly… It’s interesting knowing that I have directly impacted millions and millions in revenue - whether it was from proposal automation, analyzing phone call sentiment to improve customer relations, or improving Lead Generation workflows to get new clients. It does come with a lot of downsides too though... Watching people get laid off because I “automated” their jobs (stupid choice by businesses in my opinion. You should be ENHANCING individuals, not outright replacing them). I led teams of devs with WAY more experience than me and I'm grateful that they gave me more respect than I deserved. All this while still knowing my place though.. It was tricky to manage. I was humbled and mentored multiple times over... I just wanted to LEARN things. Now? Titles don't really matter at the end of the day. People just want to work with people that can respect the fact that we are all working towards the same goal. I am probably a junior engineer in some cases, a principal in others, or maybe at best an intern? I don’t really know. The real warning: Don’t be complacent. Stop assuming so much of AI and start reviewing the fundamentals as you learn things. I wrote this out of frustration of some peers who accepted defeat within the tech job market and also out of frustration of recent experiences in the business world... I just don't want people wasting time by making the same mistakes I made. I hate watching people stay complacent. You have an open opportunity to UPSCALE yourself but you choose to “Accept” vibe code your way to automation without first learning the fundamentals. The ones who can vibe code their way to success have years of experience under their belt to allow them to do that, whether they know it or not. Tl;dr - eat a chipotle bowl for me. I'm not summarizing this. You will likely run this through AI to summarize this and that is OK with me. As long as you put in the work at some point in time down the line.

Startups at SF: - 996 - $180k base salary with equity that never pays out - "founding engineer" with no business impact SMEs in OC: - $250k base + bonuses based on value given - impact executive decision making - build many cool apps - true unlimited PTO Why pick startups?

I turned 30 today and all of a sudden I take a 20 minute nap out of nowhere. is this it? is that the rest of life?

The biggest regret I have is credit history. I make well over six figures and am shoving each paycheck into fixing my credit and girlfriend's credit and get denied for a loan of $2,500. Now I wish I just paid back the minimal + 20% so I could have had the cash instead of paying off all debt aggressively. My credit score has barely moved up with that path, so why not just pay minimum + a healthy amount? Financial literacy for underserved people needs to be solved. 😔

Twice now that I have been burned for money this year. $15k total. both came at times in which I went above and beyond to give value only to get ghosted while watching the implementation succeed. no fucking way is the CEO of a $50m+ invested company arguing over $5k over a delivered product. no fucking way is the CEO of a $2m profit revenue company refusing to pay me $10k because they just "didn't feel like it" anymore. they're out with their family enjoying the holidays while I sit here alone for another two months trying to keep mine together. the best people to work for are the ones who value true grit, perseverance, and a passion for a shared vision.. not the ones that want to get from point A to point B without reading the details. truly, the CEOs who know what its like to stay in the grind mentality with your team going through the same sacrifices are the ones worth working for. also this might be my downfall post, but thats fine. I just wanted to protect my family.

DAY 30/30 "GET-A-JOB" CHALLENGE - FINAL RESULTS Starting work tomorrow 🙂 Signing a contract-to-hire tomorrow for two weeks for startup. Bank job hasn't reached out, but I'm hopeful they will soon and after the two weeks I can reassess my offers. Not the best 10 anniversary today for me and my girlfriend though.. she's truly a gem that I can't take for granted, so I look forward to this being the lowest point of our lives and going up from here so I can take care of her. thanks everyone for the fun 30 days. I'll write an article soon to help detail out the path I took, the lessons I learned, and what tools/strategies helped me get here. 900+ followers joined for ride and I appreciate you all ❤️ time to lock in

saving my law firm $2.5m.. or $60m? idk you decide background: - personal injury law firms sign clients on a contingency based contract at ~40% of settlement fees. - if a client doesn't like the service they're getting they can find a different lawyer at any time - most law firms don't know they're going to be subbed out until one day they get a "Substitution of Attorney" (SOA) letter from another lawyer saying "hey your clients mine now" at that point they are trying to save the client so they renegotiate with the client from 40% to 33%, maybe even 25%.. and that is ON THAT DAY that you get the letter and they can still just say no. I went back and reviewed the last few months of SOAs and we had over 200.. not great. What I did: - use Gemini to process call data for all past SOAs - found that on average most clients express concerns 30 days before we get an SOA - use Gemini to process call data for current clients - Analyze for the following, and a bit more: ---client mentions wanting to find another lawyer ---client expresses frustration on communication ---client saying they are unhappy with our service - depending on results, send an automated slack message to the case manager, ops manager, and lawyer. - one of them has to reach out to call the client on that day, minimum a personal text sent to them. - generate a "post-mortem" for the case manager that fetches all last points of communication and who on their team was responsible for events leading up to that my system caught 120 potential subouts so we called all of them and a few admitted they were looking at other law firms and they were frustrated with us. we said thank you for your patience and offered them to come in for lunch at our offices, on us. no negotiations need to be made. that was three months ago. less than 5 subouts so far and down from the average of 35 per month. from the cases we saved, some have settled totaling $60m so far. 10% of that would be $6m and to be safe I will round down to $2.5 because why not. all of this while we automate root cause of the errors of anyways so future clients don't have to go through that bad communication in the first place. the above is easy to scrap together so you can get short term wins. the real value is in building it into a workflow and automating it for alerts based on certain thresholds. to optimize costs I made a predictive model to select which cases to analyze based on a number of factors found from combining the CRM, call data, and email communications based on where they are in the lifecycle of their case. lastly, we purposefully did not automate sending out a text or phone call with an AI agent. That is good for intake, but not for personal touch relationships. we are getting a lot of clients saying they don't want an AI working their case and I've been an advocate against that for quite a while now... results may vary I guess? my bonus that I get from saving them that much money? probably a pizza party. I am so locked in for these next three months that I will be feasting in pizzas.