We help chemists change the world by building technology that understands how they work. As chemists ourselves, we know what it's like when software doesn't work the way you do, stuck with systems built by people who never spent a day at the bench. Backed by Index.
About the role
Albert already runs an advanced agentic harness in production, with fully agentic loops behind our chat experiences. You will shape where it goes next: agents that work inside every surface of Albert OS rather than a chat window, act proactively on what they see in the data model, carry goals across sessions, and run as each partner organization's own specialized scientific intelligence.
What they're looking for
- Define Albert's approach to agentic reasoning and set the technical direction of the harness, from orchestration and tool use to memory, planning, and evaluation
- Extend the harness from chat-based agents to agents that operate inside the core surfaces of Albert OS, partnering with the Director of Engineering, Core Product on the experiences that run on it
- Build the reasoning capabilities that let agents work over Albert's three-layer data model, partnering with the Knowledge Architect on how scientific intent and reasoning are represented
- Design how the harness becomes the runtime for each partner organization's own specialized scientific intelligence, learning from that organization's research and reasoning history
- Make agent behavior observable and measurable through tracing and evaluation against real scientific tasks, and use that to drive reliability at the level partners need to trust agents that act
- Stay at the frontier of what model providers and agentic frameworks make possible, and decide what Albert adopts, builds, or ignores
More about this role
Albert already runs an advanced agentic harness in production, with fully agentic loops behind our chat experiences. You will shape where it goes next: agents that work inside every surface of Albert OS rather than a chat window, act proactively on what they see in the data model, carry goals across sessions, and run as each partner organization's own specialized scientific intelligence. You will be the most senior hands-on builder on the harness and a thinking partner to Albert's engineering and product leaders on how agentic reasoning should work across the platform. You'll report to the Director of Agentic AI. In 12 months, success looks like a harness that has moved beyond the chat window and become the runtime for organization-specific scientific intelligence, evaluated against real scientific tasks and trusted enough that partners let agents act, not just suggest.
- Define Albert's approach to agentic reasoning and set the technical direction of the harness, from orchestration and tool use to memory, planning, and evaluation
- Extend the harness from chat-based agents to agents that operate inside the core surfaces of Albert OS, partnering with the Director of Engineering,...
Browse similar: AI jobs · AI startup jobs · Startup jobs · San Francisco Bay Area