Back
Member of Technical Staff, Applied Research
Full-Time · San Francisco
You will turn the Wake series into the models that run in production at the largest companies in the world. The work spans finetuning, adaptation, and evals against real-time data at scale. You will spend time inside customer environments, understanding their data, constraints, and feedback.
Sample projects include:
- • Adapt the backbone to a customer's production data and beat their incumbent system under its own latency and review constraints
- • Build an eval suite that measures how well learned detectors transfer to contexts they have never seen
- • Design the supervision for a task where labels are scarce, delayed, and censored — and prove the model learned from them
Requirements:
- • Deep formal training in statistics, mathematics, or physics, or equivalent research experience
- • Hands-on experience training models. You have written training loops that run on more than one GPU and debugged them when they broke
- • Comfort reading unfamiliar codebases, data, and customer environments
- • You care whether the model works in the world, not just on the benchmark
Nice to haves:
- • Experience finetuning or adapting foundation models against production data — events, tabular, and time series count as much as text
- • Background designing evals or supervision for problems with delayed, censored, or expensive labels
- • First-author publications at NeurIPS, ICML, ICLR, or comparable venues
- • Founder or early engineer at a zero-to-one company