About the Role
The AI Engineering Team at TRM Labs is dedicated to enabling next-generation AI applications, with a particular focus on Large Language Models (LLMs) and agentic systems. Our primary goal is to develop robust pipelines, high-performance infrastructure, and operational tooling that facilitate the deployment of AI systems with speed, safety, and scale. We manage petabyte-scale pipelines, serve models with millisecond-level latency, and provide essential observability and governance for production-ready AI. Additionally, we are actively involved in evaluating and integrating cutting-edge tools in the LLM and agent space, including open-source stacks, vector databases, evaluation frameworks, and orchestration tools, to accelerate TRM’s innovation.
The Impact You Will Have
- Architect and implement a robust agentic framework that supports tool use, context retrieval, memory, and planning.
- Build intelligent, modular agents that automate investigative tasks and augment analyst decision-making.
- Extend and scale our LLM infrastructure (e.g., OpenAI, Anthropic, local models), including prompt engineering, RAG, and evaluation loops.
- Design safe, observable, and auditable agent behaviors, ensuring reliability in high-sensitivity environments.
- Evaluate performance across metrics like reasoning, latency, success rate, and hallucination, and iterate based on user feedback and system telemetry.
- Contribute to a culture of high ownership, rapid experimentation, and ethical AI deployment.
What We’re Looking For
- Strong engineering background with deep experience in backend or systems work (Python preferred).
- Hands-on experience building with LLMs, agents, and tooling frameworks (LangChain, semantic caches, vector DBs, etc.).
- Comfort working with agentic pipelines and optimizing information flow into AI systems.
- Thoughtful approach to system design, with an eye for safety, scalability, and explainability.
- High product empathy – you care about how agents impact real users (analysts) and optimize accordingly.
- Bias toward experimentation and iteration – you’re excited to try, learn, and ship fast.
- Previous experience with knowledge graphs, task orchestration, or AI safety a plus.