Senior applied ai engineer - operate

Vi Operate puts AI agents into the daily operations of large health systems, providers, and pharma organizations. Our agents talk to patients and providers in real time over the phone, read the documents those organizations run on, and take action in the systems where the work actually lives. This role builds and delivers those agents in the field.

You will own engagements end to end. You sit with a client's operations, clinical, and IT teams to learn how the work gets done today, then design the agent that does it, build it, integrate it into their systems, and stay with it through the first months in production. This is not a demo that works in a conference room. It runs against real patients under HIPAA.

The measure of this work is whether it runs without you. An agent that performs in a pilot is a good result; one that holds up after you have moved on to the next client is the product. You may often be the only engineer in the room, and we expect you to make the call there rather than take it back to someone else.

What You'll Own

The whole agent. Discovery through production and through the first months of running it. The design, the integrations, the prompts and tools, and whatever happens when a call goes wrong at two in the morning.

The client's business logic. Every client runs their operation differently. You learn how their flow works, and you're the one who turns that into the system. A few weeks into an engagement you should know their process well enough to catch the thing they forgot to tell you.

Real-time voice. Telephony, streaming audio, speech in and speech out, latency, and sensible behavior when something fails in the middle of a live call.

Unstructured input. Getting reliable structured data out of the documents a healthcare organization actually runs on, none of which arrive clean.

The write path. Taking action inside client systems of record (CRM, EHR/EMR, practice management, claims, specialty pharmacy), including the cases where an action does not cleanly succeed.

Quality after launch. How the system gets tested, what gets measured, and how anyone finds out when it stops behaving. You set that, not the client.

Building blocks for the field. The patterns and components that make the next deployment faster than the last, plus what you learn in the field going back to Product and Platform Engineering.

What We're Looking For

Integrating LLMs into workflows. You have experience taking a language model and building it into a real workflow, where it has to handle actual inputs and hand off correctly to whatever comes next.

Real-time systems. WebSockets, streaming media, event-driven architectures, or high-throughput API services in production.

Integration with systems you don't control. CRM, EHR/EMR, claims, specialty pharmacy, or similar, including writes and not only reads. You have dealt with someone else's API on someone else's timeline.

End-to-end ownership. You have owned a whole system rather than one service inside one, and you're comfortable being accountable for something with real patients on the other end of it.

Client-facing engineering. You can run a working session with a client's technical and non-technical people and come out of it with something you can build. Then you go build it.

Python and the working stack. Strong Python plus TypeScript or Node, and experience shipping production applications.

Cloud. You don't need to hand off to a dedicated engineer. You can get what you build running at scale on AWS or GCP: containers, CI/CD, monitoring, and good security habits.

HIPAA-regulated environments. You don't need to be a compliance expert, but you understand why certain data can't be logged and how to build systems that enforce it.

Nice To Have

• Voice or telephony infrastructure at scale: Twilio, LiveKit, Pipecat, Amazon Connect, or similar
• Document AI, especially extraction from low-quality scanned input
• Healthcare or pharma domain knowledge: patient services, specialty pharmacy, pharmacovigilance, clinical operations, patient access and scheduling, or revenue cycle
• Workflow engines, rules engines, or state machine architectures
• Prior forward-deployed or professional services experience at an AI or data company