Senior / Engineer Ii, AI Lab Research Engineer

Lila Sciences Lila Sciences · AI Frontier · One Charles Park, Cambridge, MA · AI

This role focuses on building agentic AI systems for scientific discovery, including workflow/code generation and evaluation mechanisms. It requires expertise in LLMs, agent architectures, and ML frameworks, with a strong emphasis on adapting these to scientific domains. Experience with long-horizon agents, RL, and evaluation design is highly valued.

What you'd actually do

  1. Agentic AI for science: Systems that perform sequential decision‑making and multi‑step reasoning to solve domain‑specific problems.
  2. Workflow/code generation: From natural language intent to typed, executable steps for lab instruments.
  3. Evaluation & reliability: Benchmarks, test suites, and telemetry to measure capability and quantify progress toward scientific goals.

Skills

Required

  • PhD or Masters in a quantitative discipline
  • strong background in machine learning
  • one domain of science (e.g. biology or materials science)
  • Strong grasp of LLMs and agent architectures (planning, tool use, structured function calling, code generation)
  • Proficiency in modern ML frameworks (e.g., PyTorch, TensorFlow, JAX)
  • experience implementing scalable solutions for complex tasks
  • Comfort collaborating across disciplines and interfacing with simulations and real lab systems

Nice to have

  • Building long‑horizon agents or RL for control/decision‑making; experience with model‑based or offline RL.
  • Designing domain‑specific benchmarks and evaluation harnesses for complex scientific tasks.
  • Digital‑twin development, calibration, and sim‑to‑real transfer.
  • Publications or open‑source contributions in AI for science (especially publications in top-tier conferences like NeurIPS, ICML, AAAI, ICLR).

What the JD emphasized

  • LLMs and agent architectures
  • long-horizon agents
  • RL for control/decision-making

Other signals

  • Agentic AI for science
  • Workflow/code generation
  • Evaluation & reliability
  • LLMs and agent architectures
  • long-horizon agents
  • RL for control/decision-making