Co-Op, AI-Driven Equation Discovery

Lila Sciences • Cambridge, MA

9d

About The Position

The PS AI team works at the intersection of Physical Sciences R&D and AI/ML, using data from high-throughput experiments in electrochemistry, formulations, and advanced materials. As a Co-Op focused on AI-driven equation discovery, you will develop and benchmark methods that extract interpretable governing equations directly from experimental data. These equations can reveal underlying physics, make predictions explainable, and provide mechanistic building blocks for digital twins and closed-loop optimization.

You will collaborate with PS AI researchers and physical scientists to shape how Lila turns experimental data into scientific understanding at scale. The work includes building end-to-end equation discovery pipelines, benchmarking symbolic regression methods on Lila campaign data, applying equation discovery to experimental datasets, developing evaluation harnesses, and integrating discovered equations into scientific models and workflows. You will also translate open scientific questions into concrete machine-learning problems and communicate findings across research and domain science teams.

Requirements

• Pursuing a Master's or PhD in ML, CS, applied math, physics, materials science, or chemical engineering.
• Strong Python skills, with hands-on experience in PyTorch, JAX, or similar.
• Experience applying machine learning to messy, real-world scientific or experimental data.
• Solid grounding in linear algebra, calculus, and probability.
• Comfort reading and implementing methods from ML research papers.
• Ability to frame open-ended scientific questions as concrete ML tasks.
• Clear communication across research, engineering, and physical science teams.

Nice To Haves

• Working knowledge of symbolic or sparse regression methods, such as SINDy, PySR, PDE-FIND, or equivalent.
• Experience with LLM-guided or agentic symbolic regression, such as LLM-SR or SR-Scientist.
• Background in physics-informed ML, uncertainty quantification, or operator learning, including PINNs, neural ODEs, FNO, or DeepONet.
• Exposure to digital twin or closed-loop workflows, including model calibration, active learning, or Bayesian optimization.
• Domain knowledge in electrochemistry, formulations, superconductors, or materials science.

Responsibilities

• Build end-to-end equation discovery pipelines, from preprocessing through candidate generation and validation.
• Benchmark symbolic regression methods, including SINDy, PDE-FIND, PySR, and LLM-guided search, on Lila campaign data.
• Apply equation discovery to live electrochemistry, formulation, and materials datasets.
• Design evaluation harnesses that measure recovery rate, generalization, noise robustness, and physical consistency.
• Embed discovered equations into surrogate models, physics-informed models, and experimental digital twins.
• Translate open scientific questions into concrete problems with datasets, baselines, and success criteria.
• Document and present findings across PS AI and domain science teams.

What This Job Offers

• Job Type: Full-time
• Career Level: Intern