Agent Product Manager (Remote - US East Coast)

Title: Agent Product Manager (Remote - US East Coast) Location: United States At Quadrivia, we have brought together some of the smartest and most passionate people in the sector to work on an ambitious goal: help make healthcare accessible and affordable for everyone on earth, by solving the chronic imbalance between the growing demand for medical care and the constrained supply of clinical skills. While externally we are starting with creating the world's most comprehensive and customizable clinical assistant to deliver routine tasks for clinicians, internally we are pioneering a new way to run a company. Any company has a boundless series of necessary but mindless repetitive processes that take much of the time of its people. Healthcare has more of them than most, and they fall on the very people who should be caring for patients. We want to free our brilliant minds to focus on what matters most and let AI agents do the rest. That is where you come in. We're looking for an Agent Product Manager who wants to own the agents on real health system accounts, on the ground with the customer, and stay accountable for whether they perform. If you've spent the last year deep in LLM systems, prompt architecture and evaluation, orchestrating agents with Claude and the latest agentic toolchains, you know the exact line between a prototype and a production-grade system, you know US healthcare from the inside well enough to tell the difference between what a health system says it wants and how the work gets done on a Tuesday afternoon, and you want to work with an ambitious team that gives you the space to put your vision to work, you've found your team. ABOUT THE ROLE Every Quadrivia customer runs on agents: a discharge follow-up agent for one cardiology service line, an ED intake agent for a rural hospital, an appointment-reminder agent for a clinic group. Each one is configured, tested, deployed, watched in production and improved — and there will be hundreds, then thousands, across service lines and protocols. This role owns one or more account's agents end to end. You are on site with the customer, you wire the agents into their real environment, and then you own whether those agents deliver the outcome they were bought for — for as long as the account is live. It is a forward deployed role and a product ownership role in one seat, deliberately, because the handover between those two things is exactly where accountability gets lost. You will be one of a few people doing this, each carrying live health system accounts in the US and other markets. You will not run the implementation programme and you will not own the commercial relationship — those are separate roles. You own the automation and whether it performs. Our tooling is getting better fast, and the explicit expectation is that within twelve months building an agent will be easy enough that this job changes shape substantially. What we need is someone who will build hundreds of agents in the meantime, and who will turn everything they learn into the templates, test suites and tooling requirements that make their own job easier. Our end goal is that agent building and management becomes self served for our customers. If that sounds like a dead end to you, this is not your role. If it sounds like the best seat in the company, read on. WHAT YOU'LL OWN Deploy — on the customer's ground - Go on site. Sit with schedulers, care coordinators, practice managers and clinical leads, watch how the work actually happens, and translate that into agent behaviour. Expect regular US travel. - Wire agents into the customer's real environment: EHR data and workflows, phone systems and call routing, scheduling platforms, insurance and eligibility steps, clinical protocols. You will work alongside our integration engineers; you are the person who knows what the agent needs from that stack and why. - Be the AI and technical voice in the room — for the customer during deployment, and for Sales in demos, proofs of concept and scoping conversations. Build - Configure agents in Cortex (our agent platform) end to end: conversation graph (nodes, edges, transition conditions), prompts, variable extraction, escalation and confirm-then-act logic. - Build and maintain journeys — the orchestration around the agent: scheduling, retries, channel selection, call windows, escalation paths. - Define core KPIs and metrics that map to improving our customers' outcomes and goals. - Work with the broader product team improving the platform and localizing our tools to new markets and segments to drive automation. Prove - Author durable regression suites in Starfish (our testing toolkit): scenarios, personas, judges and campaigns that cover an agent's real behaviour, not just the happy path. - Reproduce any production failure as an exact scenario, A/B the fix, and regression-gate the whole suite before anything ships. - Work with Clinical and Quality so that "passing" maps to clinical safety, not just to a green test. Clinical review is a gate on every behaviour change: nobody approves their own prompt change, including you. Own the outcome - Own the agents on your accounts for as long as they are live. When behaviour is wrong you find out why, fix it, and prove the fix. You close issues when the behaviour is right, not when it is explained. - Own the numbers that say whether the automation is working — containment, resolution, escalation rate, clinical quality flags — and be the person who reports them to the customer. - Separate an agent misconfiguration from a genuine platform bug, and hand the second one to Engineering with the evidence. Scale - Carry roughly twenty agents across your accounts, not one. This role only works if you are building patterns and instantiating them, not hand-crafting every time. - Turn repeated manual work into a tooling requirement and push it to Product and Engineering — you are their most important internal customer. - Write the templates, standards and definition of done for agent quality that the rest of the company builds against. WHAT WE'RE LOOKING FOR We are not optimising for seniority. We are looking for someone unusually capable who will go and find out how something works when nobody has documented it — and who can hold their own in a health system. REQUIRED - US healthcare fluency, from the inside. You have worked with or inside US provider organisations and you understand how a health system actually runs — scheduling and referrals, the front desk and the call centre, EHR workflows, eligibility and prior authorisation, and why clinical staff trust or distrust a new system. You know the difference between what a health system says it wants and how the work gets done on a Tuesday afternoon. This is not a preference. You will be in the room with practice managers, clinical leads and health system executives, and they decide in the first ten minutes whether you understand their world. - US-based on the East Coast (Eastern timezone), and able to travel to customer sites in the US regularly. We need the overlap with both our customers and our European team, so Pacific-timezone candidates are not a fit for this seat. - 3–6 years in a role that demanded fast mastery of an unfamiliar technical domain in front of a customer: forward deployed or solutions engineering, technical product, top-tier consulting, or an early-stage company where you were the person who figured things out and then shipped them. - Demonstrated fluency with LLM systems in practice — prompt design, structured extraction, evaluation, and why agents fail. Opinions formed by having shipped something. - Comfort reading logs, dashboards, traces and configuration files, and reasoning about a system from its telemetry. SQL and scripting are expected; writing production code is not required, though you are welcome to. - Exceptional written communication. Most of your output is a specification, a test case, a root-cause explanation or a customer-facing report that someone else must act on. - Rigour about correctness in a clinical context. You will change systems that talk to patients, and you are one of the last checks before that happens. You do not need a clinical background — we have clinicians and you will work with them — but you do need to take that responsibility seriously. PREFERRED - Hands-on experience with Epic, Cerner/Oracle Health, athenahealth or similar, and with contact centre telephony. - Experience with enterprise customers and stakeholder management. - Conversational AI or voice-stack experience (ASR/TTS, LiveKit, Twilio or equivalent). - Experience building evaluation or regression-testing systems for non-deterministic software. - A second working language, Spanish above all. Mandarin is also valuable. HOW WE'LL KNOW YOU'RE RIGHT One stage of this process is a build. We give you access to Cortex — our real platform, not a sandbox exercise — along with a use-case brief and the documentation, and you build an agent and the journey around it: the conversation graph, the extraction and escalation logic, and the orchestration that decides when it calls, retries and hands off to a human. We watch how far you get, what you ask, and what you do when the documentation runs out. We care more about how you approach something you have never seen than about what you already know. It is also the fairest preview we can give you of the job itself, because this is the job. WHY QUADRIVIA - LLMs promise to automate a lot of jobs. This may be one of the largest power shifts we have seen in human history and at the fastest pace. In healthcare, this means more and better healthcare for everyone on earth. We are building towards that mission. - You own the layer where our product actually meets the patient, at the moment that layer is being defined. Implementation of AI and LLMs is the place where all the interesting things will happen over the next few years. - Mission-driven work at the intersection of AI, healthcare and enterprise technology. - Competitive compensation, equity and benefits. Quadrivia | Full-Time | Product | US-based, East Coast timezone — expect meaningful on-site time with customers HOW TO APPLY Answer these five in your application. They will take you twenty minutes and they matter more to us than your CV does. We read them first. 1. What is the last thing you shipped that a patient, or the person on a front desk, actually touched? Name the organisation, say what it did, and tell us one thing that went wrong. 2. An agent or automation you built behaved wrong in production. How did you find out, and how did you work out whether it was your configuration or a bug in the platform? 3. What does a scheduler or care coordinator actually do between 2pm and 4pm on a Tuesday? Not the process diagram. What actually happens. 4. Your LinkedIn URL. Required. Applications without one are not reviewed. 5. Anything public you would point us at — GitHub, writing, a site. Optional. Two things to confirm in your application, because they are hard requirements: 1. Are you based in the US and able to travel to customer sites there regularly? 2. Are you authorised to work in the US without sponsorship, now and ongoing? We are not in a position to sponsor or transfer visas for this role. If the answer to either is no, tell us and we will close it out honestly rather than string it along. Write these yourself. We can tell, and we would much rather read three rough sentences that name a real hospital and a real failure than a polished paragraph that names nothing. Every strong answer we have read had a proper noun and a number in it. One more thing, so nobody wastes their time: this is a full-time, forward-deployed role that owns an account’s agents for as long as that account is live. We are not hiring independent, fractional or agency consultants for it.