- Jobs
- TRM Labs
- Staff Software Engineer, Agent Engineering
Staff Software Engineer, Agent Engineering
AI Tools
Tech Stack
Agent Workflow
Architect agentic frameworks supporting tool use, context retrieval, memory, and planning. Build modular agents automating investigative tasks for analysts in blockchain analytics and financial crime prevention.
About the Role
Staff Software Engineer, Agent Engineering - US Remote
Company: TRM Labs
TRM Labs seeks a Staff Software Engineer focused on agent engineering to build next-generation AI applications. The role emphasizes Large Language Models (LLMs) and agentic systems deployed with speed, safety, and scale across petabyte-scale pipelines.
Key Responsibilities
- Architect robust agentic frameworks supporting tool use, context retrieval, memory, and planning
- Build modular agents automating investigative tasks while augmenting analyst capabilities
- Extend and scale LLM infrastructure including prompt engineering, RAG, and evaluation loops
- Design safe, observable, auditable agent behaviors for high-sensitivity environments
- Evaluate performance metrics including reasoning quality, latency, success rates, and hallucination
- Foster high ownership, rapid experimentation, and ethical AI deployment culture
Required Qualifications
- Strong backend/systems engineering background (Python preferred)
- Hands-on experience building with LLMs, agents, and tooling frameworks
- Comfort optimizing agentic pipelines and information flow
- Thoughtful system design emphasizing safety, scalability, explainability
- High product empathy focused on real user impact
- Bias toward experimentation and rapid iteration
Nice-to-Have: Knowledge graphs, task orchestration, AI safety experience
Compensation: $200,000-$275,000 base (US) plus equity plan eligibility
Location: San Francisco office; 3+ days/week on-site required
TRM is a Series C company ($220M funding) providing blockchain analytics and AI for law enforcement, financial institutions, and crypto businesses. AI fluency is baseline expected.