We believe that language models, used carefully, can make expert knowledge accessible to people who can't afford to pay for it. We're building an AI legal assistant that helps small businesses understand their contracts without hiring a lawyer at $400/hr. The NLP work is genuinely difficult — retrieval-augmented generation across long legal documents, reducing hallucinations in high-stakes answers, managing context windows with thousands of clause variations across jurisdictions. We're not looking for someone who has already solved every one of these problems. We're looking for someone who reads the latest research on a Sunday morning and gets excited about applying it to something that actually changes people's lives. Our stack is Python, LangChain, and Azure OpenAI. We use Hugging Face for evaluation and fine-tuning experiments.
Responsibilities
Build and improve the RAG pipeline powering our legal document assistant
Design and run experiments to reduce hallucination rates in long-document QA
Work with the product team to translate user feedback into NLP improvements
Write clear technical documentation for every component you own
Review AI research relevant to our domain and propose what to test next
Requirements
3+ years of production NLP engineering — deployed systems, not just notebooks
Solid experience with retrieval-augmented generation (RAG) pipelines
Hands-on with LangChain or similar orchestration frameworks
Familiarity with Hugging Face transformers for evaluation or fine-tuning
Experience deploying on Azure or a comparable cloud environment
Able to read and reason about AI research papers, not just implement tutorials
Benefits
Real mission — legal access for people who can't afford it
Fully remote across EU and US time zones
$90,000 – $115,000 base salary + equity
Conference budget and research time built into your schedule
Small team — your decisions matter and ship quickly