The role is for an engineer who can train and ship machine learning models, not only experiment with them. It calls for strong math, solid PyTorch skills, comfort writing low level code in Rust, CUDA or C, and a good grasp of how Transformers work. A strong candidate cares about getting results into a real product and is happy to work closely with customers on a brand new team.
How it fits the AI era
The work is about building the models that sit behind a product that finds problems and suggests fixes on its own. AI takes over slow, costly analysis of replays and user behavior, and can simulate users to test flows. What stays human is choosing which problems are worth solving, judging whether a model’s output is actually useful, and deciding what is good enough to ship.
How it builds AI experience
Someone in this role would train and deploy their own deep models on very large sets of real product data, including session replays, events and errors. That is hands on experience that few teams offer. To prepare, build a small project that trains a Transformer on sequence data such as click streams or page changes, and take it as far as a working demo.
Worth showing
- Rust or CUDA projects
- Strong math background
- Shipped ML to production
This is our own read, written from the employer’s posting and the company’s Spotlight. The employer’s posting is the source of truth for the role.
About PostHog
All-in-one product analytics, session replay, and feature management platform for engineering teams.
#J-18808-Ljbffr…