The company builds a product context layer. It collects and stores a product's analytics, errors, session replays and business data so people and their agents can query and act on it.
These include product analytics, replay vision, AI observability, error tracking, an MCP server for querying product and business data, and a data warehouse that runs custom SQL insights across product and customer data.
Over 500K organizations have installed it, largely through word of mouth.
The company is building a self-driving product and training its own deep ML models. A new AI Research team leads this work. It handles petabytes of events, errors, session replays and revenue data, with the goal of surfacing insights and suggesting fixes automatically.
The team is brand new, so you will have real autonomy, close contact with customers, and a strong say in what ships.
Responsibilities
Research what comes next and what performs better, then take it into production.
Improve session replay analysis. Issues are already detected in replays, but this is costly and does not scale. Using stronger models and a large volume of DOM mutation data, aim to make it faster, cheaper and best in class.
Build user behavior prediction that spots drop-offs, suggests improvements, and may lift conversion rates automatically, without heavy manual analysis or agent overhead.
Develop synthetic user testing that simulates users to catch confusing flows or breakages before release, with the aim of removing testing as a major engineering bottleneck.
Requirements
A PhD is a plus. Strong math is essential, unless you have previous experience training models, in which case strong math is still required.
Strong PyTorch skills.
Strong coding ability in Rust, CUDA or C. Rust is preferred.
Solid understanding of Transformers.
A product mindset, since you will be building things.
Any level of experience is accepted, and new graduates are welcome.
Not a fit: purely theoretical work, notebook-only work, or research aimed only at publishing. Publishing is allowed, but shipping comes first.
Conditions
This is a hybrid-remote role based out of the London office.
Tuesdays and Thursdays are meeting-free. The company works async first, favouring pull requests, then issues, then Slack.
Only US citizens or people holding a US visa can apply.
Accommodations are available during the interview process on request.