Adentris (YC P25) Is Hiring

About Adentris

Adentris (Y Combinator) builds AI-powered compliance and gross integrity infrastructure for behavioral wellness providers. Our systems publication existent diligent records, find objective and regulatory risk, and put it successful beforehand of the group who tin enactment connected it. We tally connected accumulation PHI for paying customers, truthful each creation determination reaches existent patients.

We're besides architected otherwise from astir healthcare AI companies: each exemplary conclusion runs wrong our HIPAA-compliant Azure environment. No outer LLM APIs ever touch diligent data. That data-sovereign architecture is simply a halfway logic endeavor customers take us, and you'll ain it.

The founders

You'd beryllium joining 3 founders, not conscionable a company. Dmitry Karpov (CEO) is simply a second-time YC laminitis pinch a way grounds successful B2B endeavor products and go-to-market. Sergey Yudovskiy (CPO) is besides a second-time YC founder: he antecedently ran ElectroNeek (YC W20) arsenic CEO, scaling it to ~$5M ARR crossed 30+ countries. He besides comes from a family of physicians, which is why claims denied complete archiving errors are personal, not abstract. Alex Odin (CTO) is simply a third-time laminitis who led AI astatine ManyChat ($140M+ ARR) and built user products utilized by 25M+ people; he owns the multi-agent architecture down each Adentris module.

You'll activity pinch each 3 of america daily, and your statement to a determination is 1 Slack connection long.

The role

You'll ain the method guidance of the level extremity to end. This is not a elder IC domiciled pinch a bigger title. You determine really objective information is modeled, really our AI pipelines are evaluated and trusted, and what the level looks for illustration successful 2 years. Then you build it.

You'll beryllium successful customer calls pinch Directors of Compliance, time off pinch a problem cipher has written down yet, and vessel it wrong the quarter. When there's a difficult telephone connected architecture, aliases connected whether a exemplary output is safe to show a clinician, you're successful the room making it.

This is correct for you if you've been the method halfway of gravity connected thing real, and you'd alternatively ain an ambiguous problem than a well-specified ticket. It's incorrect for you if you want a defined scope aliases a squad that already exists. You're the personification who creates them.

Your first 90 days

• First 30 days: vessel to production. Take complete 1 of our 5 unrecorded modules (Documentation QA is the busiest), get its information harness nether your control, and vessel an betterment a customer notices.
• By time 60: ain the EHR information layer. We merge pinch the behavioral-health-native EMRs (Kipu, Alleva, BestNotes and others), and the ingestion and normalization architecture crossed them is yours to set.
• By time 90: you've made astatine slightest 1 architecture determination the institution will unrecorded pinch for years, defined really we measure exemplary output earlier a clinician sees it, and group the barroom the adjacent engineering hires will beryllium measured against.

What you'll own

• Architecture crossed the afloat stack: information model, ingestion, AI pipelines, exertion layer, infrastructure
• The objective information layer: ingesting and normalizing messy EHR data: FHIR, HL7, and the galore formats that dress to beryllium them
• AI systems clinicians really trust: evaluation, crushed truth, correction analysis. Calling a exemplary is easy. Proving the output is correct is the job
• Product judgement from data: spell into existent customer data, find value gaps cipher has flagged, move them into shipped features
• The engineering bar: arsenic the squad grows, you specify really we hire, review, test, and ship

What we're looking for

• 5+ years shipping accumulation software, pinch existent ownership of systems that outlived your involvement
• Deep TypeScript/Node and Python; precocious React and Next.js
• Data engineering astatine scale: schema design, query optimization, pipelines complete ample volumes of semi-structured records
• Production LLM systems, not demos. You've built the information harness that kept them honest
• Communication that carries weight: arsenic clear pinch a clinician, a founder, and an engineer

Strong signals

• Founding aliases early technologist astatine a startup that reached existent scale
• Regulated information acquisition (healthcare, finance, legal); you've been done a SOC 2 aliases HIPAA audit alternatively than only publication astir them
• Hands-on pinch EHR information and objective formats; embeddings and retrieval successful production
• Self-hosted aliases backstage exemplary conclusion acquisition (Azure ML, vLLM, aliases similar)

Details

• Full-time, distant (US), pinch squad meetups successful San Francisco
• Competitive net positive founding-level equity
• Direct statement to the founders; short way from thought to production

Stack: TypeScript, Node.js, React, Next.js, Python, SQL/NoSQL, Azure-hosted LLM inference, Docker, Kubernetes