AI Engineer Intern/Working Student
Neurosense
Software Engineering, Data Science
Munich, Germany
Posted on Oct 9, 2025
We’re building real, production-grade AI agents—not just demos. If you’re hands-on, curious, and love exploring new agentic patterns, you’ll feel at home here.
What you’ll do- Design, build, and ship long‑running/background agents that can recover from failures, resume work, and keep state.
- Implement tool-using agents (function calling / tools API) that operate across internal services and third‑party systems.
- Make agents data‑heavy: stream, batch, and chunk large inputs; orchestrate multi‑step jobs; manage memory; persist context.
- Build evaluation + observability for agents: traces, metrics, playbooks, test harnesses, offline/online evals, A/Bs.
- Prototype fast: compare models, prompts, and planning strategies; run experiments; instrument everything.
- Collaborate with product + customers to turn messy workflows into robust agent workflows.
- Shipped at least one serious agent project (not a weekend toy). Bonus: it ran continuously, touched real data, or automated a real business process.
- Python as your main language (TypeScript a plus). Comfortable with async, APIs, and writing clean services.
- Experience with agent frameworks (e.g., LangGraph, OpenAI Agents sdk, CrewAI, Tools, custom planners) — or you’ve built your own light framework.
- Worked with RAG / vector search (pgvector, Pinecone, Weaviate, FAISS) and structured outputs(Pydantic/JSON Schema).
- Familiar with queues + schedulers (Celery, RQ, Kafka, Airflow/Prefect) and background job design.
- Practical infra: containers, CI, cloud (GCP/Azure), secrets, monitoring.
- Multi‑agent patterns, graph planners, tool routing, function‑calling best practices.
- Retrieval over large, messy corpora; compound systems that mix search, extraction, and decision‑making.
- Small team, big ownership. Ship iteratively, measure, and improve.
- Pragmatic about models and vendors; choose what works and prove it with evals.
- We care about reliability, safety, and user trust as much as speed.
Send us the GitHub link to the best agent project you’ve worked on and add:
- What problem it solves and why you built it.
- One design decision you’re proud of.
- One thing you’d improve if you had another week.