Founding Engineer (Full Stack) (US)

What you'll do

• There's no lane. You'll touch all of it. Monday. A customer's agent is struggling with a pricing objection it hasn't seen before. You dig into the conversation logs, figure out where the LLM lost the thread, and ship a fix to the orchestration layer. By lunch it's live. Wednesday. Arjun sketches out a new approach to generative UI on a whiteboard. The widget needs to adapt its layout in real-time based on what the prospect cares about. You prototype it that afternoon and have something demo-able by end of day. Thursday. The eval pipeline flags a regression. Agent quality dipped on a specific customer's demos after yesterday's deploy. You trace it to a prompt change, roll it back, and add a test so it doesn't happen again. Friday. You're building the infrastructure for agent versioning, so customers can preview changes before they go live. This is the kind of problem nobody has solved well because nobody has had to. You're designing it from scratch. The technical pillars:
• The Agent. LLM orchestration, conversation flow, tool use, voice. The agent needs to understand products it's never seen before and explain them clearly to prospects with different levels of context.
• The Experience. Embeddable, generative UI that adapts in real-time. Fast load, complex interactive demos, works across products and screen sizes.
• The Platform. Where customers configure agents, manage flows, review conversations, and measure performance.
• Evals and Self-Improvement. The pipeline that measures agent quality and feeds learnings back. How demos get smarter without manual intervention.
• Infrastructure. When a prospect hits "start demo," the agent responds instantly. No cold starts. No loading screens.

What you need

• You've built systems that real users depend on, and you've seen how they break. You can point to something in production and talk about what happened when people used it, not just how it was designed.
• You have strong engineering judgment. You know when something is ready to ship and when it isn't. You can move quickly without compromising the integrity of the system.
• You care about making LLM systems work in the real world. You understand the gap between a demo and a production system, and you're interested in closing it through evals, iteration, and system design.
• You're comfortable working across the stack when the problem requires it. Not because you're trying to be "full-stack," but because the hardest problems don't fit neatly into one layer.
• You move fast, but you don't create hidden problems. You've shipped quickly under pressure, and you've also rolled things back or killed features when they weren't working.
• You take ownership without waiting. If something is broken or unclear, you fix it. You don't need to be asked.

Why now

What this isn't

Frequently Asked Questions

Related Roles