Engineering Intern
SUMMER 2026 · 12 WEEKS
REMOTE (US) · FLEXIBLE TIMING
We don't hire interns to do intern work. We hire emerging builders and treat them as engineers.
We explicitly do not use years of experience as a hard qualifier. A 19-year-old dropout who's been building and shipping real systems for two years may have deeper evidence of obsession, ownership, and taste than a 10-year veteran in a narrow swim lane.
The intern is someone who could be our AI engineer, backend engineer, or FDE in 18 months — you just don't have the resume yet. You have side projects with real users, evidence of going deeper than anyone asked, and learning velocity that's off the charts.
What you'll do
You'll work with a mentor to scope a project in week 1, then own it end-to-end. You're a real engineer on the team — standups, production ships, code review, real deliverables. The track emerges based on your strengths and where the team needs help most.
AI track: Casey evaluation infrastructure, prompt engineering, ranking improvements, simulation tooling. Mentor: Alice.
Data platform track: Outcome data pipelines, reconciliation tooling, analytics infrastructure, integration improvements. Mentor: Greg.
Product track: Self-serve features, deployment tooling, customer-facing dashboards, internal tools. Mentor: Gaurav.
What we look for
- Obsession. Side projects with real users or open-source contributions where you went deeper than asked. Evidence of staying up because the problem was interesting, not because it was assigned.
- Technical fundamentals. Strong CS fundamentals. Comfortable in TypeScript. You use Claude Code as a force multiplier, not a crutch — you catch when the AI gives bad output and fix it.
- Learning velocity. When you encounter something you don't know, you learn quickly. Take-home tests drop you into real EMR integrations or AI scheduling datasets and expect you to make sense of them.
- Communication. The Loom walkthrough in the take-home tests evaluates whether you can explain your thinking clearly. Building without clear communication is only half useful.
The process
- 01 Intro Chat — 30 min, remote. Interviewer: Daniel. Key prompt: "Show us something you built. Not for a class — something you built because you wanted to."
- 02 Take-Home — 48–72 hr window, async. Pick one of three tracks (AI, data platform, or FDE/integration). Real data, real problems. Record a 10-minute Loom walking us through your solution.
- 03 Review + Live Extension — 60 min. Daniel plus the track mentor. First 20 min: questions about technical choices. Next 30 min: extend your take-home live, using Claude Code on your laptop.
- 04 Offer
Apply
Send a short note to [email protected] — show us something you built because you wanted to. Link, screen recording, or a paragraph describing the hardest technical problem you ran into and how you solved it.
Contact
Email: [email protected]