Machine Learning Engineer (Onchain Risk)

Role overview Build and operate models that assess whether AI agents’ onchain payments match the spending behavior their owners approved. The work combines transaction risk scoring, interpretable decisions for finance reviewers, and feedback from payment outcomes; you’ll work directly with the technical lead.

Responsibilities - Score payment requests using an agent’s past activity and information about the receiving party, with results fast enough to inform a live decision. - Detect unusual behavior that may stay within stated spending limits, such as payments to newly observed addresses or repeated amounts just below a cap. - Give finance teams clear reasons for flagged payments so they can review and decide whether to proceed. - Incorporate approval and rejection outcomes into model improvements, tracking both unnecessary holds and missed risks. - Develop a future forecasting model for agents’ monthly spending based on their transaction history.

Requirements - Write production-quality Python. - Have shipped a model used to inform financial or risk decisions, such as fraud, credit, abuse, or anomaly detection, and monitored its performance after launch. - Reside in a country where the organization can hire.

Nice to have - Experience with onchain data, transaction graph or sequence models, or real-time feature serving. - Familiarity with PyTorch, XGBoost, or LightGBM. - Experience explaining model decisions to people responsible for reviewing or overriding them.

Benefits and work setup - Salary range is $205,000–$255,000 in the US; offers elsewhere are benchmarked against the local market and paid in USD or optionally USDC. Includes 0.60% vesting equity. - Health coverage is provided, with terms varying by location. Other benefits include 25 vacation days plus public holidays, a year-end closure, 14 weeks of paid parental leave, a laptop and workspace allowance, and an annual learning budget. - Flexible hours, with a shared availability window from 14:00 to 17:00 UTC; outside that period, hours can be arranged flexibly. The role may be combined with another job if there is no conflict.