Careers
Build the reliability layer for AI training
We are a small team of systems engineers turning a decade of research into the software that keeps the world's largest training runs alive. Every role below ships code that runs inside production training jobs.
Open roles
6 positionsSenior Systems Engineer, Record & Replay
Own the recording and replay engine: API interception, determinism enforcement across ranks, and verified-state merging.
GPU Runtime Engineer
Bring Watcher's GPU extension to production: CUDA and NCCL recording, per-device health signals, and automatic recovery on real clusters.
Memory Analysis Engineer, XScope
Extend XScope's tensor-level attribution, retention diagnosis, and preflight projection to new parallelism strategies and frameworks.
Software Engineer, Queryable Debugging
Build the query layer on top of partial replay: state, last write, call history, value history, and joins across threads and steps.
ML Infrastructure Engineer, Integrations
Make Teyon work with every launcher and framework customers run: torchrun, Slurm, Kubernetes, FSDP, DeepSpeed, Megatron-LM.
Founding Solutions Engineer
Be the engineer in the room with design partners: run technical evaluations, measure savings, and build the monthly savings report customers see.