Founding AI Engineer

About the role

**About ViyaMD**

ViyaMD gives attorneys and claims professionals physician-level clinical analysis of medical records: the depth of reasoning that drives case outcomes and claims decisions, delivered in minutes instead of weeks.

Our founding team is physicians, attorneys, and AI engineers from Cedars-Sinai, UCLA Law, the University of Michigan, Carnegie Mellon, Microsoft, and Cruise.

**The Role**

Attorneys ask questions that don't have tidy answers. *When was this symptom first documented? Did the delay in diagnosis change the outcome? What did the treating physician know on the day of discharge?*

Each of those becomes a feature. And each one starts as a prototype that works beautifully on three cases — then has to survive four thousand pages of scanned charts, contradictory notes, and records that arrive out of order.

That distance is the job. You'll take intelligence surfaces from the first version to dependable: medical chronology, clinical insights, deposition analysis, drafting, chat. Build the agents behind them, dig into real cases to find where they break, fix what you find, and prove it stayed fixed.

This is a founding engineering role. You'll own surfaces end to end rather than tickets, work closely with the founding team and our Head of AI on where to push next, and help set how this team builds as it grows.

**What You'll Own**

• **Intelligence surfaces, end to end.** You take a hard clinical-legal question and turn it into a feature attorneys rely on — from prototype through the version that holds up at volume.
• **The agents behind them.** Prompts, tools, and multi-step reasoning, tuned until the system is reliable rather than merely impressive.
• **Grounded output.** Every claim traceable to the page it came from. Confidently wrong is the failure mode that matters here.
• **Knowing when it's actually working.** Digging into real clinical data to find failure modes, building the evaluations that measure them, and using what those reveal to improve prompts, agents, and models.
• **Clinical fidelity.** Translating clinical concepts, terminology, and standards of care into criteria a system can be measured against.

**What We're Looking For**

• You've shipped LLM-powered features that survived contact with messy real-world data, not just demos.
• You measure rigorously. You define the metric, build the pipeline, and improve the system from what it tells you — rather than shipping on vibes.
• You're comfortable with clinical material. A medical degree isn't required; real familiarity with clinical concepts and terminology is.
• You write the production code and build your own analysis tooling. Python preferred.
• You're comfortable working hands-on with PHI in a HIPAA-regulated environment.
• You move fast with incomplete information and figure out what's worth building next.

**Nice to Have**

• Agent harnesses, tool-using systems, or fine-tuning LLMs
• Medical-legal work, or citation-heavy systems where being wrong has consequences
• Structured extraction at document scale — OCR, layout, labs, billing
• Running evaluations against real production traffic

**Compensation and Benefits*

• Competitive base salary, commensurate with experience
• Meaningful early-stage equity
• Medical, dental, and vision