ML engineer job hunt takeaways

I interviewed at Bloomberg, FB, Stripe, TikTok for MLE positions. Went to onsite for three of them, failed one after the first round. It was a demoralizing and at times random process. There seems to be no standardized interview process since every company has their own product focusing on different ML problems. Here are some tips I wish I knew before starting.

1. Have good breadth on ML fundamentals. Know how to derive the gradient updates for all basic algorithms. Understand the intuitions behind the math so that you don't have to memorize as much. Don't focus just on supervised learning—also study unsupervised stuff like clustering.

2. Research what the company's product is and read their tech blog posts. Recommendation systems? Review collaborative filtering and newer deep learning models. Fraud detection? Brush up on how to train classifiers on unbalanced data. This will pay off big time in system design rounds and might be the most important thing I learned after failing onsite rounds.

3. Understand the data cleaning, model deployment, and evaluation steps of the ML pipeline. Companies want to know that you can deal with messy, real-world data and know how to debug models efficiently.

4. While you're focusing on ML prep, also solve enough LC questions. The coding bar is at least as high as for SWE positions. It's always better to focus on general algorithms and problem-solving approaches instead of grinding away at hundreds of problems. (You probably don't have the time.)

Feel free to add more.

#machinelearning #engineering #bloomberg #stripe #bytedance