- San Francisco, California, United States
- On-site
- Full-time
- $140K–$300K per year
- 2–5 years
Design, train and deploy machine learning models for real-time risk calls in a payments setting, owning data pipelines, evaluation, and metrics.
Compensation
Base salary: USD 140,000 to 300,000 per year
Responsibilities
- Own the data pipelines and modeling design for core models, including a small language model used for contextual understanding.
- Convert plain-language user approvals into structured, checkable limits covering vendors, amounts, items and time windows.
- Score automated actions in real time to flag drift, substitutions, injection attempts and overspending before funds move.
- Make risk calls from identity, user history, device and vendor signals, choosing whether to approve, request confirmation or refuse.
- Own evaluation, labeling and metrics, tracking false approvals and false blocks as product measures.