LLVM Research Group Research Assistant - Verified AI Code Generation

Description, Responsibilities, and Qualifications

The LLVM Research Group, led by Professor Vikram Adve, is looking for a graduate research assistant at 50% FTE starting in Summer 2026 with the possibility of continuing into Fall 2026 and Spring 2027.
 

This position will work on our ongoing project to develop techniques for verifying the correctness of LLM-generated code. So far, we have developed an example system for this for the Ansible programming language (described in this pre-print) and are looking to develop such a system for Bash. We are looking for an RA to help with the development of such a system, including through the use of LLMs to generate parts of this system, such as formal specifications of the behaviors of POSIX utilities, and developing techniques to validate the generated system.

A good candidate will possess the following skills:

• Familiarity with using LLM code-generation and optionally LLM training and inference
• Interest and/or experience with code testing, experience with test generation is a plus
• Familiarity with *nix (Linux, Unix, or other similar OS) utilities and systems programming
• Highly motivated and willing to learn
• Able to work with other students in a team
• Good communication skills, including writing and presentation skills

The following additional skills would be a strong plus, but are not required:

• Familiarity with agentic LLM code-generation systems
• Familiarity with programming language theory and compilers
• Good programming skills in OCaml