PhD Position: Hierarchical Reinforcement Learning for AI for Mathematics

About the role

Four-year fully funded PhD at Chalmers’ Division of Data Science and AI, embedded in Chalmers’ formal mathematics programme and funded through WASP, Sweden’s national AI initiative. The project applies hierarchical reinforcement learning (HRL) to automate proof generation and conjecture formation inside interactive theorem provers such as Lean or Isabelle. The research combines AI, formal methods, NLP, and cognitive science. Starting salary is SEK 35,725 per month. Deadline: 18 June 2026.

Responsibilities

• Design and implement HRL algorithms for automated theorem proving.
• Integrate learned policies with symbolic world models inside interactive proof assistants.
• Publish at premier AI venues including NeurIPS, ICML, and AAAI.
• Complete advanced courses through the WASP Graduate School.
• Contribute up to 20% undergraduate teaching.

Requirements

• Master’s degree in Computer Science, Mathematics, or a closely related field.
• Strong foundation in theoretical computer science, foundational machine learning, or mathematics.
• Proficient in Python and at least one compiled language (C, C++, or Java).
• Solid understanding of reinforcement learning theory.
• Fluent English, written and spoken.
• Physical presence in Gothenburg required; valid residence permit by start date.

Nice to have

• Experience with formal logic, proof assistants (Lean, Isabelle, Coq), or theorem proving.
• Familiarity with transformer architectures or language model fine-tuning.
• Prior research output or participation in competitive programming.

How to apply

Apply through the Chalmers online recruitment portal. Submit CV, cover letter, transcripts, and two referee contacts. The reference number is REF 2026-0066. Deadline is 18 June 2026.