Founding AI-Native Engineer

Rockstar is recruiting for an AI-native operations platform for patient access — the behind-the-scenes work of getting patients enrolled in assistance programs, verifying coverage, and moving cases through review. Their thesis is simple: AI handles the routine, humans handle the judgment calls. They're in build mode for their first deployment, a patient access infrastructure company serving emerging biopharma, with go-live targeted later in 2026.

They're hiring one or two senior engineers to build this platform with them. Given the size of the team, each hire carries the weight of a founding engineer — you will own entire verticals of the product, make architectural calls that we live with for years, and ship code that real patient cases flow through starting this fall.

They are looking for engineers who build with LLMs the way previous generations built with databases — as a first-class architectural primitive, not a bolt-on.

You won't be handed tickets. You'll own problem spaces end-to-end. That looks like:

• Designing and shipping the AI agents that run patient case lifecycle — application intake, document extraction, data validation, coverage verification, patient communication, missing-info follow-up — including the evaluation harness, confidence thresholds, and human-in-the-loop boundaries
• PHI-aware data architecture: de-identification pipeline that strips identifiers before inference, the tokenization layer that maps results back, and the audit trail that logs every model call
• Integrations with healthcare systems of record and dependent services – as well as the orchestration layer that coordinates across them
• The production bar: on-call, incident response, rollback strategy, and strict uptime SLAs
• Working directly with operations, IT, and clinical teams during discovery, UAT, and go-live — you'll be in the room, not behind a PM

• 5+ years shipping production software, with meaningful time owning backend systems that mattered when they broke
• Deep expertise in at least one modern backend language (Python, TypeScript, Kotlin, Go, or similar) and the judgment to know when the choice matters and when it doesn't - they are still finalizing their stack, and you'll help decide
• Comfort across the full backend picture: relational databases, cloud infrastructure (AWS preferred), containerized services, API design, and the operational discipline that comes with running regulated software
• A real sense for UX — you care how the thing feels to use, not just whether it works. You've sat with users, sweated the details, and pushed back on designs that looked good but didn't hold up in the workflow. Formal design training isn't required, but familiarity with Figma, design language, or having partnered closely with designers is a plus
• You've built something real with LLMs — orchestration, retrieval, tool use, structured output, evals, handling model failure modes. You have informed opinions about when to reach for an agent vs. a pipeline vs. a plain function, and you can defend them
• Track record of taking an ambiguous problem, shaping it into an architecture, and shipping it
• You treat model behavior as something to test, measure, and govern. You've thought about what "evals" means for a system where being 95% right isn't good enough
• Bonus: healthcare or other regulated-industry experience; prior early-stage or founding-engineer time; Salesforce ecosystem; eligibility engines, document extraction, health compliance or case-management workflows

Eastern Time preferred, +/- 3 hours.

It's not a research role. It's not a prompt-tuning role. It's not a "wrap an API and call it an agent" role. It's also not a role for someone who needs a lot of scaffolding around them — they are small, the customer is counting on them, and the timeline is real.