AI Internship: Self-Hosted LLMs for Data Migration

In this 6-month internship starting in Spring 2027, you'll test whether sovereign, locally-run AI can reliably scope complex legacy data migrations so our clients can get accurate fixed-price offers.

Your Tasks

• Benchmark open-source LLMs that can be self-hosted on sovereign infrastructure against leading cloud models on three migration-scoping tasks: understanding schemas, extracting business rules and assessing data quality.
• Deploy and run candidate models locally with tools such as Ollama, vLLM or Hugging Face, and see how sovereign AI works in practice.
• Design and build a prototype agentic pipeline that analyzes legacy schemas, code and sample data on its own.
• Uncover hidden business rules in legacy systems, including .NET-based applications, that often only show up after in-depth analysis.
• Validate data consistency and quality, and assess the complexity of a representative legacy migration case.
• Generate automated effort and complexity estimates, along with step-by-step migration guidelines, from a locally-run model.
• Define criteria and best practices for when a local model is accurate enough to replace manual scoping in fixed-price migration bids.
• Evaluate and document the resource, cost and licensing trade-offs between self-hosted and cloud models.
• Shape how we scope one of our most frequently fixed-price-quoted services for clients with strict data sovereignty needs.

What We Offer

• A dynamic, collaborative workplace with a highly motivated, multicultural team working across international sites
• Monthly AI meetups with our expert AI team, where you can ask questions and move your topic forward
• 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

• You have a Bachelor or Master degree (EPFL, ETH, University) in Computer Science or related discipline
• You are comfortable programming in Python and SQL, and ideally have some familiarity with .NET-based systems
• You have worked with LLMs and, ideally, tried deploying them locally (e.g. Ollama, vLLM, Hugging Face)
• You are curious about agentic AI and keen to build and benchmark your own pipeline
• You have a basic understanding of database schemas and data quality
• You like to work in a structured way and to share your findings clearly
• You are fluent in written and spoken English, as well as German or French

Application: Please attach your academic transcripts to your application. Applications without transcripts will not be reviewed.