Founding AI Engineer Intern Full Time

Founding AI Engineer Intern — PillPilot (Summer 2026)

PillPilot is building the AI-native automation platform for pharmacies. Pharmacies run on software from the 1990s, and the people who use it spend their days on the phone, in faxes, and typing prescriptions by hand. We’re replacing all of it with agents.

There are three of us. You’d be the fourth. We work out of an office in Manhattan, and our software runs in real pharmacies today.

The job is whatever the company needs that week, but mostly two things: writing code that ships into our actual products, and figuring out problems nobody has scoped yet. You’ll get briefs like “our agent’s drug-entry accuracy dropped on insulin prescriptions this week, figure out why.“ If you’d rather work off a backlog of tickets, this isn’t the right job.

What you’d actually do

• Voice agents — Own pieces of our AI voice calling stack. Prompt architecture, tool calls, latency tuning, eval pipelines.
• Agent orchestration — Build tool-based agentic workflows on top of our PMS integrations (PioneerRx, BestRx, PrimeRx).
• Evals & routing — Own the eval harness for our data entry and voice AI. Define what “good” looks like across accuracy, quality, and latency, then drive real decisions about prompts, models, and routing.
• Data entry automation — Define edge cases, build test scenarios from real prescriptions, and stress-test the agent before it touches a live pharmacy.
• Internal tooling — Trace viewers, eval dashboards, demo modes. Whatever lets us debug faster.

You might be a good fit if

• You’ve built something with LLMs that actually had to work. A multi-step agent, tool use, a voice or chat product. You can tell us what broke and what you did about it, in detail.
• You’ve shipped things people use, and you can point to product decisions you made, not just code you wrote.
• You’re fast with AI coding tools. They’re a real multiplier for you, not a thing you’ve tried a couple of times. You’re an expert at using Claude Code to get things done.
• You notice broken things and fix them. The intern who emails to ask if they should take initiative is not the intern we want.
• You want to work at a startup. The hours are uneven and the office is in person, five days a week.

Nice to have, none required

• Familiarity with healthcare data.
• Voice or realtime LLM tooling (Retell, Vapi, Twilio, OpenAI Realtime, LiveKit).
• You’ve built an evals pipeline that caught something before it went out.
• You’ve started something — a side project, a company, a GitHub repo with stars you didn’t ask for.

Logistics

• $5,000/month plus NYC housing
• In-person, five days a week in Manhattan
• Summer 2026
• Currently enrolled in a degree program, or returning to one in fall 2026
• US work authorization required. We can’t sponsor visas.