Role Overview
Working student (15–20 hrs/week) at Thorit GmbH in Böblingen, Germany. Build a model-agnostic LangChain and LangGraph foundation for internal AI tools, migrate existing agents onto it, and help colleagues build consistently on the shared stack. Hybrid, mostly on-site.
Responsibilities
- Map internal LLM tool usage: models, calls, and data flows
- Build the internal LangChain/LangGraph reference stack: model-agnostic LLM layer, tool calling, MCP, structured outputs
- Migrate existing tools and agents onto the shared stack
- Compare LLMs with evals and tracing; recommend models per use case
- Create templates and docs for consistent tool building, including Lovable apps
- Connect the stack via APIs, webhooks, and a shared data model
Requirements
- Enrolled in computer science, software engineering, AI, data science, or business informatics
- At least one self-built LangChain or LangGraph project (code required)
- Hands-on with multiple LLM providers and their trade-offs
- Solid Python; first experience with REST APIs and SQL
- Very good English; German is a plus
- Live close enough to Böblingen for at least one office day per week
Nice to Have
- LangSmith or other LLM tracing/evals, RAG, vector stores, MCP
- TypeScript with Lovable, Supabase, or Postgres
- Interest in writing a thesis on applied LLM or agent topics
Benefits
- Real ownership; work goes live from month one
- Flexible hours around lectures and exams; more hours in semester breaks
- Paid vacation pro rata (31 days full-time equivalent)
- Access to frontier LLMs and senior engineer code reviews
- MacBook, modern Böblingen office, lunch subsidy, snacks and drinks
- Thesis option and path to Junior AI Engineer role after graduation