Equation Discovery Agent

WHAT IT DOES

Given experimental or observational data, discovers the governing differential equations or algebraic relationships. LLM proposes candidate equation forms based on physical intuition and domain knowledge, symbolic regression fits parameters, evaluator checks fit quality and parsimony (Occam's razor — simpler equations preferred). Iterates with evolutionary search: mutating, combining, and refining candidates. Applications: physics (discovering conservation laws), biology (population dynamics), engineering (system identification), finance (regime dynamics). Pairs LLM creativity with automated verification.

WORKFLOW

1. Ingest data
2. LLM proposes candidate equations
3. Symbolic regression fits parameters
4. Evaluate fit + parsimony
5. Evolutionary iteration (mutate, combine, refine)
6. Verify on holdout data
7. Report discovered equation + confidence + physical interpretation

SKILLS

INTEGRATIONS