Transform your career with expert guidance designed for ML engineers. Whether you’re breaking into the field, acing interviews, or climbing the career ladder, MLEPath provides actionable strategies and proven tools to help you succeed, grow, and thrive in the dynamic world of AI/ML.
DeepSeek is shaking up the AI landscape, but its real impact might be in the engineering behind it.
Let’s talk:
🧠 Sparse Mixture of Experts
🧠 Multi-head Latent Attention
🧠 Model Distillation
Are these the future of AI models? I’m joining the team at Rotational this Friday, Feb 7 at 12PM ET to break it all down.
🚀 Join us live on Zoom!
🔗 https://buff.ly/4b6g7FD
Got pushback on my last video—some claim seniority is all that matters in ML hiring. That’s completely false.
Here’s what actually gets you in the door:
🔹 Pedigree – A Stanford undergrad, a Bengio lab paper, FAANG experience.
🔹 Referrals – The closer the connection, the stronger the signal.
🔹 What you can do – The right frameworks, a project people actually use.
These are the three biggest hiring signals in industry.
#MachineLearning #TechCareers #Hiring
Neural networks are said to be ‘inspired by biology,’ but their learning process couldn’t be more different.
Imagine if, as a newborn, you were given 80% of all human knowledge—shuffled randomly, with no logical progression—and then never taught anything again.
That’s how we train AI today. Maybe it’s time to rethink our approach?
ML demos shouldn’t be smoke and mirrors—they should drive real AI progress. Too many teams fake it with flashy demos instead of solving real problems. A good demo cascade builds on past work, aligns with production constraints, and reduces uncertainty. Let’s stop selling AI and start building it.
🔗 https://buff.ly/40Z0HPX #MachineLearning #AIResearch
Just heard from an M2 candidate I’ve been working with—after waiting 3 MONTHS, they finally got a FAANG offer.
Not all ghosting means no… but also, no one should have to wait that long for an answer.
🎙️ ML engineers, you don’t want to miss this!
Kevin Van Horn (30+ years in ML, ex-Adobe Sr. Staff) unpacks Bayesian probability, ML optimization, and the hidden challenges in real-world AI. Get practical insights & deep knowledge—listen now: https://buff.ly/3Cz4Hx7
#MachineLearning #TechPodcast
Early in your ML career, every decision feels irreversible. But the best engineers don’t aim for perfection—they build with reversibility in mind.
Understanding the difference between one-way and two-way doors will help you iterate faster and build better.
#Meta warns that is will fire leakers in a leaked memo
Now that's meta.
https://buff.ly/4aBqUqV
Had a great conversation with a Senior Staff ML candidate in my office hours yesterday—it reinforced something important:
At higher levels, ML System Design interviews require a deep dive. If you don’t structure your time right, you’ll run out before getting to the details that matter.
Here’s how I recommend structuring your time.
By the way, I host office hours every week, sign up link in Substack email each Monday at 3PM Pacific. First come, first served.
📍 Sign up: https://mlepath.substack.com
Preparing for a Meta ML System Design interview? Their questions often focus on just 3 key areas:
➜ Recommender Systems – Ranking & personalization
➜ Harmful Content Detection – Filtering spam & misinformation
➜ Topic Modeling (NLP) – Structuring massive text data
Focusing your prep on these areas will make a huge difference. I help candidates clear this round every month—if you need a mock interview, I’m here: https://mockdesignround.com 🚀
Breaking into open source isn’t as hard as you think. Dr. Rebecca Bilbro’s advice:
➜ Work with OSS libraries
➜ Find issues as you use them
➜ Fix them & submit PRs
That tiny script you always write? Others need it too—contribute it upstream.
Listen to more from Rebecca: https://open.spotify.com/show/0KuQkv8pouA2SR5nHzRX2B
My first big launch at Adobe went smoothly—until the next morning.
Production was down. My change broke everything. My teammates were exhausted. I was horrified.
But I owned it, fixed it fast, and my director trusted me more because of it. Failure isn’t a career-ender—it’s a lesson.
You’re not getting into big tech without #LeetCode—but grinding endless problems isn’t the solution.
➜ Understand what companies actually test
➜ Solve problems strategically, not blindly
➜ Part 1 of 4: MLE coding interviews broken down
Full breakdown: https://buff.ly/4jyu7LZ
NeurIPS 2024 felt different—less academic, more implementation.
LLMs alone won’t cut it. Reinforcement learning, smarter architectures, and real-world benchmarking are where the field is going next.
The future of ML is taking shape.
The ‘throw more data at it’ era is ending. OpenAI is struggling for breakthroughs, and when you plateau, others catch up fast.
This is great news for ML—we need diverse techniques, not just one trick that made a few billionaires.
AI can help with interview prep—but don’t rely on it blindly.
➜ Use AI to refresh what you already know
➜ Ask specific questions with context
➜ Always ensure you understand the why behind the answer
AI should assist, not replace, real learning.
AI-generated content has a distinct bullet-point spacing, especially from ChatGPT. Once you notice it, you start seeing AI slop everywhere.
Are we all paying more attention to this now?
Want to break into ML engineering? The real path looks like this:
➜ Build a project for yourself—see if you enjoy it
➜ Fill in gaps with courses (now you know what’s missing)
➜ Build something useful for others
➜ Start applying & prepping
Actively building matters more than just studying.
When applying for MLE jobs:
➡ Stay organized—batch applications by similar requirements
➡ Prep for specific interview types
➡ Expand filters only if needed
Too much variation = burnout. Stay focused and sane!
@paco I agree with you, the hype merchants' vision of current state of LLMs is extremely unrealistic and reminds me of Hinton saying "no more radiologists in 5 years" back in 2016
Still, I've been in the industry since 2011 and must admit that the progress we have made in NLP is remarkable. Not on the "AGI in 20 years" scale, but on "AGI in 200 years" scale.