Pincites is an AI contract negotiation assistant embedded in Microsoft Word. We help legal teams review and redline contracts faster. You can read more at pincites.com.

We’re an early stage startup (YC S23) with great investors and early traction. Our customers include tech unicorns and organizations you know. We’ve doubled our revenue in the first three months of 2024 and are at an inflection point in our growth.

About our team

Our founders combine experience from big law, big tech, and high-growth startups.

• Sona is a practiced attorney with expertise in transactional law and intellectual property. She has a background as a strategic consultant in legal tech, where she helped companies automate their internal processes for almost three years.
• Mariam worked as a software engineer at Meta and as a senior PM at GitHub, where she led teams building security products.

We’re looking for a sharp, ambitious machine learning engineer fluent in building AI-native products — someone who knows how to turn messy real-world data into performant models, fine-tune and deploy LLMs, and design feedback loops that make AI systems learn continuously.

At Pincites, you’ll help transform our negotiation data into fine-tuned models that power the next generation of AI contract review. You’ll lead the evolution of our core intelligence layer — from prompt-based heuristics to data-driven models — and help define how legal negotiation knowledge becomes scalable, repeatable, and self-improving.

About Pincites

Pincites is building an AI-native contract negotiation platform for legal and procurement teams. Our product lives inside Microsoft Word and helps teams negotiate faster and more consistently. It learns how top companies — like Ramp and Vercel — negotiate today, then automates that workflow with AI-generated redlines and comments tailored to their playbook.

Backed by Nat Friedman, General Catalyst, Liquid 2 Ventures, and Y Combinator, we’ve built strong traction with enterprise legal teams globally and are on track toward building a billion-dollar company. We’re seed-stage, fully remote, and assembling a world-class team.

About the Role

You’ll design and build the systems that make Pincites truly intelligent:

• Convert our 32K+ playbook “checks” into structured training datasets
• Fine-tune LLMs for clause classification, redline generation, and comment writing
• Build pipelines to capture feedback from human reviewers and feed it back into models
• Collaborate with product and backend engineers to deploy models behind our API
• Evaluate performance and reliability — moving from prompt-engineering to robust inference

You’ll be hands-on across the full ML lifecycle: data → model → evaluation → deployment.

Who You Are

• You have 3–10 years of experience building production-grade ML or AI systems
• Strong in Python, PyTorch, and modern ML tooling (Hugging Face, Weights & Biases, OpenAI fine-tuning APIs)
• Deep understanding of LLMs, embeddings, RAG, and fine-tuning (LoRA, adapters, or custom heads)
• Experience building or maintaining data pipelines and labeling systems
• Can ship backend integrations (Go or TypeScript familiarity a plus)
• Excited by the challenge of turning unstructured legal data into usable, scalable AI
• Thrive in ambiguity, move fast, and enjoy owning problems end-to-end

Bonus:

• Experience in legal tech, document intelligence, or compliance AI
• Familiarity with pgvector, GCP, or serverless infrastructure

Why Join

• Turn a massive, real-world dataset into a competitive AI moat
• Work directly with founders from Meta, GitHub, and top law firms
• Ship models that go straight into customer hands — visible impact, zero bureaucracy
• Competitive salary, meaningful equity, and full remote flexibility

Anyone can build an AI-powered demo. The technical challenges come when you try to build a product customers love.

• We push the limits of what the AI can do. Getting the best out of LLMs requires orchestrating complicated chains of requests using prompts that have been methodically developed. We benchmark and iterate our prompts to achieve results others can't match.
• We integrate with our users' current tools. For lawyers that means Microsoft Word, and we're not afraid of becoming experts in technologies like OOXML or officejs to better serve them. It's the difference between a good experience and a great one.
• We sweat the details to build a great product. We keep things fast with hundred of parallel request to OpenAI, keep them intuitive with constant iteration of our frontend, and keep them reliable by fixing every bug customers experience.

To power the above we keep it simple with our tech stack:

• React / TypeScript for our web frontend and Word add-in
• Go for our backend API
• PostgreSQL with pgvector for our data store
• GCP and Vercel for our backend and frontend hosting
• OpenAI as our foundation model provider