Formal Verification Scientist (Lean 4 & Mathlib)

Formal proof research

Convert rigorous mathematical arguments into machine-verifiable Lean 4 proofs. The work investigates what current proof assistants can represent and automate, while contributing to research on AI formal reasoning.

• Build clear, correct formalizations in Lean 4 or related proof systems.
• Analyze general and specialized arguments for hidden assumptions, missing steps, and separable formal components.
• Develop formal proofs in areas where automation struggles or fails.
• Help AI researchers design, test, and improve formal verification methods.
• Write readable, reproducible scripts that follow mathematical practices and Lean conventions.
• Advise on decomposing proofs, selecting lemmas, and organizing formal structures.
• Explain prover failures precisely, including complexity, unavailable lemmas, and library constraints.
• Compare formalized classical proofs with their textbook presentations.
• Construct proofs that reveal patterns or generalizations implicit in the original mathematics.

Required background

A master's degree or higher in Mathematics, Logic, Theoretical Computer Science, or a closely related field is required. You need a strong foundation in rigorous proof construction across areas such as algebra, analysis, topology, logic, and discrete mathematics.

Practical experience with Lean 3, Lean 4, Coq, Isabelle/HOL, Agda, or a comparable system is essential, with Lean 4 strongly preferred. You should translate informal mathematics into organized formal proofs independently and be comfortable working where mathematics and computer science meet.

Useful experience

Type theory, the Curry–Howard correspondence, and proof automation are relevant. Contributions to Mathlib or other large formalization projects are helpful, as is experience with theorem proving that requires substantial manual support.

Data annotation, quality evaluation, or AI training experience is welcome. Clear communication helps explain formalization decisions, difficult cases, and reasoning strategies to colleagues from different disciplines.

Contract arrangement

The hourly contract is fully remote with flexible scheduling for 10 to 40 hours weekly. Assignments provide exposure to advanced AI training and the role of formal mathematics in verifying model reasoning.

Further projects may support ongoing collaboration or contract extension.