Co-op, LLMs for Decision Making

Your Impact at LILA

Lila Sciences builds AI systems that accelerate discovery across the physical and life sciences. Within Physical Sciences AI, our decision making efforts develop the algorithms that drive experimental decision-making, closing the loop between models, experiments, and the next thing to try. We're now exploring how large language models can extend that capability: encoding domain priors, proposing candidates, reasoning over campaign history, and pairing naturally with established algorithms like Bayesian optimization for sample-efficient search.

As an LLMs for Decision Making Co-Op, you will work at the intersection of LLMs and Bayesian optimization, prototyping and evaluating approaches that combine language model reasoning with principled experimental design. Your work will land in the decision making stack that powers experimental campaigns across Lila's AI Science Facilities.

What You'll Be Building

• Contribute to LLM-based decision-making methods for experimental campaigns, focused on a well-defined sub-problem
• Prototype approaches that combine LLM reasoning with Bayesian optimization, active learning, or design of experiments, with mentor guidance
• Build evaluation frameworks that benchmark LLM-augmented strategies against established Bayesian baselines
• Help integrate promising methods into the decision making stack used across physical sciences campaigns
• Document findings and share results through write-ups, presentations, or contributions to internal libraries

What You'll Need to Succeed

• Pursuing a Master's or PhD in Machine Learning, Computer Science, Statistics, Applied Mathematics, Physics, Chemistry, Materials Science, or a related quantitative field
• Strong programming skills in Python and familiarity with ML frameworks such as PyTorch, JAX, or similar
• Foundation in Bayesian methods, Bayesian optimization, or probabilistic modeling
• Experience with large language models including fine-tuning, test-time compute, and benchmarking in applied settings
• Ability to turn open-ended scientific decision-making questions into concrete ML tasks with clear baselines and metrics
• Comfort iterating on experiments and analyzing results in research-style codebases
• Clear communication and interest in collaborating across ML and physical science teams

Bonus Points For

• Experience with active learning, design of experiments, multi-objective optimization, or batch Bayesian optimization in scientific problem settings
• Familiarity with agentic frameworks and structured-output techniques for scientific reasoning
• Exposure to physical science applications such as materials, chemistry, catalysis, batteries, electrochemistry, or related domains
• Prior work pairing LLMs with optimization, planning, or decision making processes