You are familiar with how modern AI products are built and shipped. You have worked with LLMs, agents, tool use, and the standard moving parts, and you have opinions about what works and what does not.
You have shipped real software to real users. You can describe in detail something you built, what broke, and how you fixed it. You move across the stack and have product instincts. You do not need a clean spec to start.
If you bring deep experience in any of the following, tell us. We are hiring across a few different shapes of this role, and these are the areas where we want strong owners.
- Building AI agents that operate over large or messy datasets, with thoughtful sampling, summarization, and bounded queries.
- Designing multi-agent systems and pipelines. Layered agents, generator and critic loops, planners and executors, and orchestration where one agent's output drives another's work.
- Agentic chat that reasons over real customer data and can act on it.
- Building AI products that ask very little of the user. This means the UI itself, plus the scaffolding behind it: smart defaults, agent harnesses, and prompt structures that let the agent do the heavy lifting on the user's behalf.
- Evals, observability, and the feedback loops that turn customer feedback into measurable product improvements.
- Voice agents of any kind. Outbound or inbound calling, real-time conversational systems, coaching, or anything else where audio is in the LLM loop.
- Workflow or pipeline builders where users compose actions through chat.
Stack
TypeScript across the stack, React and Vite on the frontend. Postgres with Prisma. AWS infrastructure. LLMs from Bedrock and OpenAI. We build our own agent code rather than using a framework.