AI / LLM Engineering (Werkstudent)

About the jobCognitX AI GmbHis building a privacy-first AI advisor for data analytics. We let people ask complex questions about their data in natural language and get trustworthyanswers, visualizations, and automated actions. Our architecture connects directly to enterprise sources, leverages a knowledge graph for context, and executes analyses through a secure compute layer. On-prem and cloud deployments are first-class.The RoleAs a Working Student in AI/LLM Engineering, you’ll help design, build, and evaluate Agentic AI systems that turn questions into reliable analytics and automated workflows. You’ll work across model prompting/tool-use, data access, and evaluation, collaborating closely with our founders on real enterprise use cases.TasksWhat you’ll do:Build and improve agentic workflows (tool/function calling, planning, self-checks) for analytics, summaries, and visualizations, task automation.Implement adapters/tools to connect LLMs with internal and external services.Contribute to our FastAPI backend (clean interfaces, validation with Pydantic, tests).Develop evaluation metrics to measure accuracy, latency, and cost.Optimize prompts, retrieval/contexting, and execution strategies for privacy, reliability, and performance.Ship in containers (Docker) and collaborate on deploys (Kubernetes), CI, and observability.Document decisions and share learnings with the team.RequirementsWhat you bring:Experience with LLMs/AI agents (function/tool calling, RAG, MCP, or agent frameworks) and Machine Learning fundamentals.Strong Python skills for production code and data work (typing, tests, packaging).Familiarity with APIs and microservices (FastAPI preferred), Git, and containerized dev (Docker).Solid problem-solving, clear communication, and a proactive, ownership mindset.Currently enrolled in Computer Science, Data Science, AI, or a related field (eligible to work as a Werkstudent:in).Nice to have:Retrieval/RAG stacks (embeddings, chunking, evaluators), Redis, PostgreSQL.Kubernetes, GitLab CI, observability (logs/metrics/traces).React/TypeScript for light UI tooling.Interest in EU privacy/GDPR and safety-by-design for enterprise AI.Exposure to self-hosted/open models (e.g., Llama, Mistral etc) and model serving.Knowledge-graph concepts; graph queries.BenefitsWhy CognitXWork on meaningful, privacy-first AI with real enterprise pilots.Learn fast with tight mentorship from a team of Senior AI Experts.Flexible hours, remote-friendly, and impact from day one.Competitive Werkstudent compensation.