🌍 Location: Munich (office-first) ⏰ Start: immediate 💰 Package: Competitive compensation 🗣️ Languages: German + English

We're building a nature foundation model - trained on proprietary IoT sensor data, satellite feeds, and bioindicators to learn a unified representation of ecosystem state. You'll start with owning the Insight Chat, the natural-language layer land managers use to query and act.

Nature has never been queryable. For the first time, a developer in solar and wind, a farmer, or a conservation officer can ask a living ecosystem a direct question - is this site recovering? which species moved in? is this land compliant? what can I do to increase resilience against heatwaves? - and get an answer grounded in what our sensors actually collected in the field and all the collective knowledge that is stored away in scientific papers, domain databases and often the heads of ecologists. We're building the interface between human decisions and the planet's own data. Get it right and capital starts flowing toward the ecosystems that need it, at scale. The stakes are real, and that's what makes it worth doing.

What you'll actually work on

• Retrieval over heterogeneous data: Fusing time-series acoustic detections, geospatial layers, and structured biodiversity records into a retrieval layer that stays traceable back to source.
• Grounding + evaluation infrastructure: Developing benchmarks to judge answer quality, latency and cost budgets, regression suites on real customer queries. You'll build the evaluation loop to improve the our reasoning layer week over week.
• Guardrails that matter: Answers feed into compliance workflows, so explainability and citation-back-to-data are first-class, not afterthoughts.
• Agentic workflows: Moving past single-turn Q&A toward multi-step reasoning over a site's full data history → the "predictive maintenance for the planet" thesis, in code.
• End-to-end ownership: data pipeline → retrieval → LLM integration → the frontend the customer actually touches.

You'll fit if you

• 3-5 years of full-stack product engineering experience.
• Have strong frontend & backend skills (You should be fluent in web technologies both on frontend and backend. Having Infra / IaC experience is a bonus.)
• Hands-on experience building data products and LLM-powered features in production, e.g. RAG pipelines, prompt engineering, working with vector databases, and evaluating/monitoring model output quality.
• Have a product mindset and taste: you think about the user first, not just the feature. You have developed mental models that guide you in a data constrained environment.
• Are comfortable owning features across the tech-stack without hand-holding.
• Are AI-pilled: you use these tools daily and have strong opinions about why most AI products are mid.
• Care that the thing is fast, correct, and traceable, and get twitchy when it isn't.
• Have experience in an early-stage startup.

As a person: