36 Labs is an early stage startup using AI and machine learning to build creativity into LLMs.
We're looking for an AI Engineer to own and push forward the agentic pipelines at the core of what we're building. Joining at this stage means real ownership over your work and the opportunity to establish yourself as a core member of a fast-growing startup.
What you'll be doing
- Building and improving agentic LLM pipelines: multi-agent orchestration, tool use, context engineering, structured generation
- Running experiments with new models and techniques, and taking what works to production
- Closing the loop between what the system generates and how it performs in the real world, so it improves every cycle
- Treating prompts, knowledge bases, and agent reasoning as first-class engineering artifacts: versioned, tested, continuously improved
- Turning raw, messy real-world data into structured signal that drives what the system generates
- Implementing scalable data pipelines, optimizing models for performance and accuracy, and ensuring they are production-ready
- Building evals and pipeline observability so we can tell what's working, catch regressions, and measure output quality
What we're looking for
- Production experience building LLM-powered systems: prompting, context engineering, RAG, agent architectures, evals
- Strong programming skills with proficiency in Python and experience building production applications, including version control, CI/CD, and modern software development practices
- Statistical intuition and technical judgment: you recognise when a model's output is unreliable despite appearing convincing, know when to use traditional deterministic approaches, and understand how to optimise an LLM
- Comfortable moving fast where things are loosely defined and priorities compete; you wear several hats and own problems end-to-end
- Strong technical communication: able to translate complex AI concepts into architectural decisions and actionable implementation plans
- Can showcase solving hard problems with artifacts like past projects or GitHub contributions