Who we are
Sapien is the engine behind the world’s most important companies. We build agents that learn how a business works from its financial and operational data: its language, operations, and economics. That understanding helps people find opportunities, make better decisions, and act.
We work with complex businesses, from manufacturers and restaurant groups to the Fortune 500. Teams at companies including Bayer, Cooper Standard, Blink Charging, Carlex, and Odeko use Sapien to understand their businesses and improve how they run. Our ambition is to build the self-optimizing company by making its people and systems dramatically more capable.
We’re headquartered by Madison Square Park in NYC, with $15M+ raised and backing from General Catalyst, Neo, and operators from OpenAI, Google, Microsoft, Ramp, and Stripe. We’re bringing together AI researchers, engineers, and operators to build intelligence for the people who build everything else.
The role
You'll build the AI agent capabilities that power Sapien's financial and operational intelligence. This means designing novel architectures for reasoning over complex financial and operational data, implementing verifiable and observable agent workflows, and building systems that learn and adapt to each company's unique operations.
This is a research-meets-product role. You'll work on cutting-edge agent capabilities—from observability and library learning to semantic search and multi-modal parsing—and ship them directly into production for customers.
What you'll do
Design and implement agent architectures that enable observability, human-in-the-loop verification, and precise context control across complex financial and operational workflows.
Build library learning systems that reduce LLM dependencies by learning reusable patterns for planning, code generation, and data localization from customer interactions.
Create graph-based company representations and develop efficient search methods using embeddings, semantic clustering, and custom retrieval strategies.
Build multi-modal parsers that unify diverse financial and operational data sources (Excel, ERPs, CRMs) into coherent, queryable schemas that agents can reason over.
Design benchmarking and evaluation suites that quantify Sapien's accuracy, reliability, and business impact across different customer workflows.
What we're looking for
Strong algorithmic thinking. Demonstrated through ML research, competitive programming, mathematics, or building novel systems from scratch.
Experience with modern agent frameworks, LLMs, and AI systems: fine-tuning, retrieval augmentation, tool use, or agentic architectures.
Comfort working end-to-end: from implementing research ideas and prototyping architectures to deploying production systems and iterating on real customer feedback.
How you work (values we care about)
Adaptability: You thrive on diverse, open-ended problems. One day you're implementing a new agent architecture, the next you're debugging a customer data issue—and you bring the same intensity to both.
Ownership: You take problems from research paper to production. You don't wait for direction—you identify what needs to happen and drive it to completion.
Opinionated: You have strong technical perspectives on what works and what doesn't. You push back on approaches you disagree with and make the team's thinking sharper.
Mission-driven: You're energized by building AI that transforms how businesses run. You connect technical decisions to real customer outcomes and care deeply about the impact.
Collaborative excellence: You share knowledge, give thoughtful feedback, and invest in making the team better. You bring new ideas from papers and discussions to elevate everyone's thinking.