Build a Safer World.
TRM Labs provides AI-powered intelligence solutions that help public and private sector agencies investigate and disrupt crime. TRM's platforms enable investigators to trace illicit activity, build cases, and construct operating pictures of threat networks. Leading agencies and businesses worldwide rely on TRM to make the world safer and more secure.
TRM Labs provides AI-powered intelligence solutions, including our AI investigations platform Ardence, that help public and private sector agencies investigate and disrupt crime and global security threats. At TRM, you'll join a mission-driven, fast-paced team made up of experts in law enforcement, data science, engineering, and financial intelligence, tackling complex global challenges daily.
As an AI Scientist on our new AI Research team, you will build the models and agents behind the intelligence that powers TRM's products. Part of that is training foundation models on blockchain data, one of our most differentiated data layers, and part is building autonomous AI agents that generate intelligence at a scale hand-built heuristics cannot reach. You'll work on Intel Cortex and other ambitious, cross-cutting ideas, with a broad view across intelligence rather than a single risk category. This is pioneering work that does not exist elsewhere in our industry, and it directly drives customer outcomes, from seizing illicit funds to disrupting criminal networks.
The impact you'll have here:
- Train ML models on blockchain data, including foundation models, and take them from research to production.
- Build AI agents (and fine-tune LLMs where it helps) so the intelligence underneath TRM's products is generated by models and agents, not only hand-built heuristics.
- Find new ways to generate net new intelligence, using patterns across all intelligence layers (Network / On-chain / Protocol) that competitors can't replicate.
- Ship production-quality code: ML models, agents and data pipelines, with rigorous evaluation and scalable architecture.
- Take on ambitious, high-impact bets that raise the ceiling of what our intelligence can do, and drive measurable outcomes: billions in illicit funds traced, seized, or prevented.
What we’re looking for:
- Quantitative background (Computer Science, Statistics, Engineering, Mathematics, Physics). Advanced degree or equivalent research experience is a plus.
- Strong machine learning foundations, built over years of applied ML and/or research (modeling, evaluation, experimentation). You don't need to have trained a foundation model or built agents before, but you need the fundamentals to do it.
- Strong problem-solving mindset: gets to the root cause, leverages creativity, and thrives in ambiguity with a hypothesis-driven, iterative approach.
- 5+ years of hands-on machine learning experience, with strong Python, SQL, and AI-assisted coding tools (Cursor, Claude Code).
- Experience with distributed data systems (Spark, BigQuery, Airflow) and shipping production ML code.
- Strong async communication (Loom, Slack, Notion), strong technical communication in English, and ability to own work independently in a remote environment.
- Bonus: experience with LLMs, AI agents or fine-tuning; domain curiosity or prior work in blockchain, crypto, fraud, money laundering or financial crime.
About the Team:
- AI Research is a new team focused on ambitious, cross-cutting ideas in attribution and intelligence.
- Fully remote and globally distributed. Most overlap occurs 7am–12pm PST for meetings and collaboration.
- All team members must have at least 4-hour overlap with PST business hours.
- You'll work in partnership with the category pods, whose expert-built heuristics provide the high-confidence labels our models learn from.
Learn about TRM Speed in this position:
- Rapid Customer Impact. Deliver last-minute, high-stakes projects for key customers — turning around attribution insights in days, not weeks.
- MVP to Production. Build complex algorithms with tight timeframes (1-2 sprints for MVP) and iterate based on real customer feedback.
- Parallel Execution. Juggle multiple priorities across categories and stakeholders while maintaining quality and speed.
