Job Description

Start date: February 2027 · Duration: 6 months

Please include your latest academic transcripts with your application. Applications submitted without transcripts will not be considered.

AI coding assistants are evolving from simple autocomplete into autonomous agents that plan, edit multi-file codebases, run tests and iterate on their own output. Alongside proprietary tools such as Claude Code, a fast-growing ecosystem of open-source coding agents and harnesses (OpenCode, Pi, OMP and others) pairs open LLMs with different orchestration, tool-use and context-management strategies.

In this 6-month internship starting in February 2027, you will investigate how far open-source agents can go: can a small open-source model paired with a carefully tuned, custom harness match or beat a large model running behind a generic harness such as Claude Code?

Your tasks

• Set up and benchmark open-source coding harnesses (OpenCode, Pi, OMP) against Claude Code across a shared suite of coding tasks
• Design a rigorous benchmark protocol: task tiers, success criteria and scoring (accuracy, code quality, turns, cost, latency)
• Test whether small open-source models with a tuned, custom harness can match or beat large models behind a generic harness in specific situations
• Analyze failure modes and identify which harness-design levers matter most (planning, tool use, context management, error recovery)
• Extend harness design to other tasks: project management, visualization, data analysis
• Write up your findings in a technical report, present results and demos internally, and create reusable templates and guidelines for open-source coding agents

What we offer

• A dynamic, collaborative workplace with a highly motivated, multicultural team working across international sites
• The chance to make a difference in people's lives by building innovative solutions
• Internal coding events such as Hackathons and Brownbag sessions, plus our technical blog
• Monthly After-Works organized at each location

About your profile

• Comfortable with Python
• Genuine interest in LLMs, autonomous coding agents and developer tooling
• Rigorous and curious, comfortable designing experiments and interpreting data
• Good written communication skills (benchmark reports, internal documentation)