What are ML engineering roles like these days?

I love this question! Here's my experience after interviewing ~20 companies and working at a few in the last 2 years

Classic ML

This is your using XGBoost to build some prediction models for specific tasks. Example: notification personalization at ecommerce company. A lot of the skill here is really data engineering with spark, and understanding classical models. Feature engineering

Deep Learning

At about 4-5 companies (FAANG, newly IPO'ed unicorns, late stage startups) a lot of the work was using the tons of data they had to train deep learning models + build training pipelines, checkpointing all that for transformer based models for ad ranking, using sensor data to generate health predictions + some companies where their whole thing is building ML models for a very specific use case e.g. ML for heart issue diagnostics

LLM companies

A lot of new startups are fine tuning LLMs for very specific use cases that are unlocked by LLM. Eg. document parsing for vertical X

Moonshot research companies

These are research engineering roles for extremely competitive companies working on tough problems where the reward is really high once you crack it. E.g. LLM optimization, RL models for markets, Distributed LLM inference with privacy preserving, robotics, AGI labs