Member of Technical Staff, Full-Stack/Front-end

TypeSafe AI•San Francisco, CA

9d•$150,000 - $250,000•Onsite

About The Position

TypeSafe is an AI lab developing System One Models, combining the reasoning of large language models with the reliability of code to build trustworthy AI for production. The company describes itself as a small, fast team backed by an $870 million Series A led by a16z. The Member of Technical Staff will build interfaces, datasets, and evaluation tooling that advance model intelligence and support production-ready reliability. The role combines product sense, data science, and engineering. The primary technology stack is Python, with TypeScript, Next.js, and Tailwind CSS for frontend work and Kubernetes for orchestration. The company permits engineers to use helpful tools, including Claude Code and Cursor. The team works fully in person at its San Francisco office near Embarcadero station. It emphasizes collaboration, craftsmanship, emotional honesty, kindness, and impactful work. Compensation is a base salary of $150,000–$250,000 plus equity, based on leveling.

Requirements

Solid full-stack engineering fundamentals and strong coding ability.
Product intuition.
High attention to detail and a strong bar for quality.
Interest in how systems are used in the real world, beyond the code.
Comfort with ambiguity and ability to find creative solutions in scrappy, fast-moving problem spaces.
Hands-on experience implementing LLMs and understanding their capabilities and limitations.
Willingness to critically assess actual model generations rather than trust LLMs blindly.

Responsibilities

Create high-leverage datasets, products, and user interfaces that unlock capabilities and use cases for the company’s model.
Build internal data-tooling and evaluation interfaces for model evaluation, debugging, and data inspection.
Develop and own evaluation frameworks to measure quality, reliability, and emergent characteristics across model iterations.
Run analyses and experiments to understand how data, training choices, and targeted interventions affect model behavior.
Turn raw model outputs and data into clear, inspectable user interfaces, owning a domain end-to-end.
Partner with product managers and customers to translate real-world needs into model and system improvements.
Continuously iterate and explore techniques for building trustworthy AI products.