Get live statistics and analysis of Ravi | ML Engineer's profile on X / Twitter

If you really want to get into Machine Learning, it is non-negotiable that you learn both traditional ML and newer LLM stuff Before when I saw job descriptions that asked for it - I just thought it was due to the hype But now I can say for sure, you can't make it without it

Being a Data Scientist is hard Being a Data Engineer is hard Choose your hard. But after you choose, drop it and be a Machine Learning Engineer instead.

Rookie Data Scientist: "I made a model in a jupyter notebook, check it out!" Pro Data Scientist: "I solved X problem with Y model by manually collecting Z dataset. I put it in a nice web app so you can play with it - here's the link, let me know what you think."

Machine Learning Interview Prep is wild. You have to prep for: - Leetcode (DS&A or SQL) - ML Coding Questions - ML Conceptual Questions - ML System Design Questions - Behavioral Interview Questions And now potentially even LLM questions, regardless of if it's in the job or not. That's why it's important to specialize in some way, so you get clarity on what to study.

Transitioning from freelancing to full-time in ML is wild. Marketing goes from being a social media game to resume game. Sales goes from discovery calls to interviewing skills. Not a ton of resources on ML specific interview either, though they aren't as common anyway so does it even matter? Honestly going the other way is probably easier.

Is there anything more boring than a data analytics job?

Job Alert: Recruiter reached out to me with some jobs - thought I'd share here too. Here's the list: 1.- Founding Member of Technical Staff — Back-end / Systems Engineering 📍 New York, NY | 🏢 Onsite 2.- UI / Swift Engineer 📍 San Mateo, CA | 🏢 Hybrid 3.- Senior Software Engineer 📍 New York or San Francisco | 🏢 Onsite 4.- Senior/Staff Software Engineer (C++) 📍 Remote (US, Canada) | 💻 Remote 5.- Senior iOS Engineer 📍 Berkeley, CA | 🏢 Hybrid 6.- Senior Software Engineer, Core Platform 📍 San Francisco or New York | 🏢 SF/NYC only 7.- Software Engineer, Product Engineering (Full Stack) 📍 San Francisco or New York | 🏢 SF/NYC only 8.- Senior Software Engineer (Full-Stack) 📍 New York | 🏢 Onsite 9.- Software Engineer, Distributed Systems (Core) 📍 North America | 💻 Remote 10.- Senior Software Engineer (FullStack) 📍 SoHo, New York | 🏢 Onsite If you're interested reach out here: https://www.linkedin[dot]com/posts/activity-7302810476520853507-jEFS/?utm_source=share&utm_medium=member_desktop&rcm=ACoAAEAW7P4B49ZiWwUF1vtVVR6GhOinzyKlWn4

Being a Data Scientist is hard Being a Data Engineer is hard Choose your hard. But after you choose, drop it and be a Machine Learning Engineer instead.

My parents still want me to get a "safe" government job. Lol

A New Unconventional Path to Break into Tech We all know the classic catch-22 of "how do I get a job without experience, but how do I get experience without a job". Well, here's a different approach to getting around this I've been thinking about. The strategy: - Start a service-based business outside of tech - Secure and serve real clients in your chosen field - Find tedious or repetitive tasks in your service offering - Now go build a software to alleviate these pains of yours - And viola, you just got paid to solve a real problem with software Freelancing works just as well but personally I think doing a bit of both would be much better and you'd be getting paid to learn - effectively creating your own junior job. I'd say this works for whatever field you're interested in, so if you want to get into data, software, AI, or whatever, just tailor whatever you build to that field (ask chatgpt for ideas). How could you be not hirable after this? Or maybe you never even need a job. Just my thoughts of the day, let me know what you think.

Best way to learn fast: Watch a crash course at 2x speed just to get a high-level overview. Then come up with a project idea around it, and only re-watch the parts you need to build it. Finally, repeat.

My new content strategy: - 2-3 tweets per day - 1 LinkedIn Post Mon, Wed, and Fri - 2 Blog Posts per week - 2 YouTube videos per week, on the blog posts Repurposing content is the key, but let's see how it goes.

"Keep trying, you just need someone to give you a chance" is total cope

Freelancing Lesson: Everyone wants AI, but nobody wants to deal with crappy data. I’ve seen this pattern repeat again and again: ⚫ Unreliable Data Sourcing: Data is scattered, schemas are inconsistent, and nothing flows reliably. A model on top? Good luck. ⚫ Labeling gaps: supervised learning is king - but nobody is prepared for the grind of labeling. ⚫ Quality issues: Duplicates, nulls, silent errors… accuracy dies long before the model gets blamed. As they always say in ML. Garbage in. Garbage out.

Successful Machine Learning Product in 3 Bullets: - Quality data - Validated demand - XGBoost

My parents still want me to get a "safe" government job. Lol

Being a Data Scientist is hard Being a Data Engineer is hard Choose your hard. But after you choose, drop it and be a Machine Learning Engineer instead.

Machine Learning Interview Prep is wild. You have to prep for: - Leetcode (DS&A or SQL) - ML Coding Questions - ML Conceptual Questions - ML System Design Questions - Behavioral Interview Questions And now potentially even LLM questions, regardless of if it's in the job or not. That's why it's important to specialize in some way, so you get clarity on what to study.

Is there anything more boring than a data analytics job?

If you really want to get into Machine Learning, it is non-negotiable that you learn both traditional ML and newer LLM stuff Before when I saw job descriptions that asked for it - I just thought it was due to the hype But now I can say for sure, you can't make it without it

The Future of Data Science Work Where is this field headed? Was thinking about this lately. We're drowning in Gen AI hype right now, but at the same time there are ML leaders saying MLOps is the most useful skill. I think it's only natural to expect the next in demand skill to be LLMOps. The perfect blend between the two. These days we're still in a "proof of concept" phase where we're still experimenting with adding LLMs to their workflows. Once we have this figured out, we will be much more focused on operationalizing things - i.e. LLMOps. Well, that's at least my opinion, or just thought of the day. Things change so fast, who knows what will happen.

𝗢𝗻𝗲-𝘀𝗶𝘇𝗲-𝗳𝗶𝘁𝘀-𝗮𝗹𝗹 𝗠𝗟 𝗺𝗼𝗱𝗲𝗹𝘀 𝗱𝗼𝗻'𝘁 𝗮𝗹𝘄𝗮𝘆𝘀 𝘄𝗼𝗿𝗸. Here are the 3 most common reasons to go the custom route: 1. Competitive edge through proprietary data. 2. Smoother transition to scale with your company. 3. Easier to improve user experience with customized features. Custom solutions offer more flexibility that sometimes you just need. Don't settle for generic when you can have bespoke.

My LinkedIn profile just got wiped/hacked randomly. All my information and content is gone, just a blank profile now. Is this a good or bad thing?

If you want to become a machine learning engineer, Please don't lower yourself to being a data analyst. Absolute waste of time. And you risk dying of boredom.

Imagine if we could get all of our inner dialogue thoughts into text format. Then do sentiment analysis on it, what would yours look like?

Hot take: coding is easy. Well idk if this is really a hot take, but in my opinion, the hard part isn't writing code but rather the reasoning behind the code. Especially in ML, where projects are pretty complex and performance depends so much on the data. You really gotta know why you're writing that code for any model to be meaningful.

A New Unconventional Path to Break into Tech We all know the classic catch-22 of "how do I get a job without experience, but how do I get experience without a job". Well, here's a different approach to getting around this I've been thinking about. The strategy: - Start a service-based business outside of tech - Secure and serve real clients in your chosen field - Find tedious or repetitive tasks in your service offering - Now go build a software to alleviate these pains of yours - And viola, you just got paid to solve a real problem with software Freelancing works just as well but personally I think doing a bit of both would be much better and you'd be getting paid to learn - effectively creating your own junior job. I'd say this works for whatever field you're interested in, so if you want to get into data, software, AI, or whatever, just tailor whatever you build to that field (ask chatgpt for ideas). How could you be not hirable after this? Or maybe you never even need a job. Just my thoughts of the day, let me know what you think.

LLMOps is just MLOps with extra baggage. But that baggage is heavier than you think. Always remember that: 🔹 Prompt engineering never stops Traditional ML has its optimization cycles, but LLMs need constant prompt iteration. You might find yourself spending more time tweaking prompts than you ever did tuning hyperparameters. 🔹 Monitoring gets existential How do you track drift when your model's output is... creativity? Traditional accuracy metrics certainly won't help when you're model decides to write poetry instead of answering questions. 🔹 RAG is search with extra steps Retrieval-Augmented Generation sounds revolutionary, but it's really just "let me Google that for my AI." Context engineering is the real challenge. The hard part is never the tech stack - it's that LLMs break every assumption you have about how ML systems should behave predictably. Your model worked yesterday? Cool. Today it decided to be philosophical.

Rookie Data Scientist: "I made a model in a jupyter notebook, check it out!" Pro Data Scientist: "I solved X problem with Y model by manually collecting Z dataset. I put it in a nice web app so you can play with it - here's the link, let me know what you think."

My new content strategy: - 2-3 tweets per day - 1 LinkedIn Post Mon, Wed, and Fri - 2 Blog Posts per week - 2 YouTube videos per week, on the blog posts Repurposing content is the key, but let's see how it goes.

I created an AI Lara Acosta What I mean by this is I took her LinkedIn Content Generation strategy and automated it with CrewAI and ChatGPT It follows this Agentic Framework in the image, and using it you can create your LinkedIn posts of the week in seconds. You can find it here: https://ravinderrai[.]com/linkedin-automation/ But here's how it works in detail 🧵