FOUNDING RESEARCH ENGINEER

TL;DR

• • Build the context infra powering the next generation of AI agents
• • Own core systems across memory, personalization, and permissioned action
• • Work with top agent teams and LLM researchers at the edge of what's possible

ABOUT COBROWSER

CoBrowser is building the trust layer for the agent-native web—the secure infrastructure that lets AI agents act on a user's behalf with permission, memory, and control.

Our first product, Kontext, gives agents persistent memory and scoped access to user data. It's already powering real-world agents in productivity, search, and consumer experiences—improving retention, user feedback, and long-term UX.

We're backed by a16z CSX and founded by operators from a16z, BCG, and the Ethereum Foundation. We're re-architecting the web so agents can safely do things for people, not just talk to them.

THE ROLE

This is not a traditional engineering role. You'll work at the bleeding edge of LLM agent infra—where system design meets behavioral optimization.

As our Founding Research Engineer, you'll design and ship core primitives for agent context, including reflective memory, session identity, and scoped permissions. You'll explore emerging areas like GEPA, DSPy, and self-distillation, and contribute to making agent behavior more controllable, trustworthy, and aligned.

You'll go deep with users—from debugging production agents to co-designing personalized memory structures with customers. Expect to contribute to open-source libraries, infrastructure APIs, and deeply technical customer deployments.

WHAT YOU'LL DO

• • Design and build memory and personalization primitives for LLM agents
• • Implement and evolve core auth/context APIs (TypeScript + Go)
• • Prototype new approaches to agent memory
• • Work closely with users building real agents (MVPs and production)
• • Push the boundaries of real-time context injection and prompt scaffolding
• • Own performance, reliability, and security of mission-critical infrastructure

YOUR BACKGROUND LOOKS SOMETHING LIKE THIS

• • 2+ years building systems at the infra or applied ML level
• • Strong TypeScript proficiency; comfortable with Go or willing to learn
• • You know how agents work under the hood: CoT, action graphs, memory modules
• • You've worked on auth/session flows (OAuth, cookie/session/token semantics)
• • You thrive in messy, high-context early-stage environments and can ship fast
• • You're curious about agent architectures and skeptical of benchmark inflation

EVEN BETTER IF YOU:

• • Built or contributed to DSPy, Letta, LangGraph, or similar agent frameworks
• • Have worked on RAG and RL infrastructure (e.g. embeddings, user-specific context stores)
• • Like clean abstractions, fast experiments, and crisp reasoning about design tradeoffs

COMPENSATION

$150K – $200K + Meaningful Equity
Location: Munich or San Francisco (in-office)
Visa support available