Founding Applied AI Lead

This job post is for our NYC location. For our Toronto location job post, go here.

We are flexible on the precise job title. It could include "forward deployed," "engineer," "technical writer," "implementation lead," etc.


Paxos Health is a Seed-stage healthcare AI startup that has raised >$6M in venture capital funding, with AI agents already operating in production with customers. Our founding team comes from Stanford, Meta, Microsoft, Medtronic, Columbia, University of Waterloo, and has founded previous startups and healthcare nonprofits.

We are building the AI-native operating system for getting patients access to the medical technologies their doctors prescribe. Too often, even when a physician decides a patient needs a healthcare product (cancer tests, genetic diagnostics, prosthetics, heart valves, neurostimulation devices, home medical equipment, etc.), health insurance companies deny coverage, and breakthrough technologies struggle to reach the people they were built to help.

We help healthcare manufacturers fight for coverage patient-by-patient. Before LLMs, this work was too difficult to use AI technology for, and human teams were forced to cut corners to get through all of the patient cases. But now, AI can help manufacturers meet increasingly complex insurer requirements and get more patients access to care.

Our progress so far:

• Five medtech and lab diagnostics companies as customers, with 40+ companies in pipeline. $182K in recognized revenue, with line of sight to $1M ARR by late summer
• Raised >$6M across two VC rounds, including our recently closed Seed
• Won the Stanford Impact Founder and IDIF fellowships. Check out the Stanford article about us and Haley's LOWkeynotes talk.

So far, we've deployed LLMs for written documentation workflows and voice AI, and we want to expand those and build out computer-use agents as well.


About the role

Paxos Health is looking for a hands-on Founding Applied AI Lead. You'll work directly with the founders and help define the applied AI function at a seed-stage company before the playbook exists. (Note that the job title itself is flexible based on candidate preferences.)

The challenge is turning probabilistic LLM behavior into reliable, auditable workflows that customers trust in production. We've succeeded with the customers we've had so far, and we want to scale our AI to many more.


This role is deeply hands-on, but the core output is not production backend code. It is the written and structured system design layer around our AI: precise agent instructions, workflow specs, output schemas, implementation playbooks, and decisions about how messy customer processes should be translated into reliable product behavior.

As an early employee, there is a significant opportunity to grow the scope of your role. For example, you could become more product-adjacent, or more engineering-adjacent, depending on your interests. (See the AI-generated diagram directly below.)

We are open to many backgrounds for this role, not just software engineering. For example, a strong senior technical writer can be a great fit for this role. Much of the work is about precise instructions, structured outputs, reusable specs, and customer-facing implementation logic. If you have owned complex documentation systems, API or developer docs, information architecture, docs-as-code workflows, or cross-functional technical content, this role applies those same skills to AI systems in production.

Responsibilities

You will own the applied-AI layer of Paxos's customer workflows. In practice, your work will include:

• Designing AI workflows from messy real-world processes: translate customer reimbursement workflows into clear system behavior, including inputs, outputs, decision rules, edge cases, escalation paths, and human review steps.
• Writing and improving agent instructions: create prompt/instruction sets that help agents extract clinical facts, apply payer policies, cite evidence, identify missing information, and generate reviewer-ready prior authorization or appeal content.
• Creating reusable implementation playbooks: turn customer-specific learnings into clear internal specs, templates, QA checklists, reviewer guidance, and implementation patterns that help Paxos scale beyond one-off customer builds.
• Defining structured outputs: design schemas, evidence tables, decision trees, missing-information checklists, and reviewer-facing summaries that make AI outputs reliable, auditable, and easy to evaluate.
• Building evals and QA processes: create test cases, eval sets, quality rubrics, and failure-review loops so we can measure and improve workflow performance over time.
• Debugging AI workflow failures: inspect outputs, identify why an agent missed evidence or followed the wrong logic, and turn those findings into improvements to prompts, schemas, workflow design, or product behavior.
• Working directly with customers during implementation: clarify requirements, understand edge cases, validate outputs, and turn customer-specific learnings into reusable Paxos capabilities.
• Helping shape Paxos's AI roadmap: contribute to our strategy for document workflows, voice AI, computer-use agents, automated prompt-writing, and other applied-AI systems.

For example, you might take a payer policy and customer cases, turn them into a structured decision workflow, write the agent instructions, design the output schema, build evals, review failures, and work with engineering to productionize the workflow.

This is not a pure software engineering role. If you have a software engineering background, you can write some production code, but that will not be a majority of the role.

Requirements

We also welcome candidates with many more years of experience than these minimum requirements, and the role can scale up in scope and higher within the compensation range.

• Minimum 2+ years of work experience in any one of the following (wide variety of fields): forward-deployed engineering, technical writing, software engineering, AI operations, legal work, software QA at AI companies, or related work.
  
  • We're considering multiple backgrounds because this is a role that the industry is only starting to define. We care most about finding someone excited about it rather than any one particular path.

• Extreme detail-orientation: you notice edge cases, inconsistencies, missing requirements, and quality issues that others overlook.
• Communication skills: strong English written and verbal communication. Comfortable interacting directly with enterprise customers.
• Strong interest in Applied AI: you use LLMs regularly and are excited to apply them to complex real-world workflows. You do not need to have been an AI engineer, but you should be genuinely motivated by AI.
• High ownership: comfortable operating in a fast-moving startup environment where processes are still being built.
• Based near New York City or Toronto, or willing to relocate. This is an in-person / hybrid (minimum 3 days per week in the office) role with the Paxos team.

You do not need to have been an AI/ML engineer, or production software engineer. You also do not need deep healthcare experience. The most important skills are precision, structured thinking, customer empathy, comfort with ambiguity, and the desire to build AI workflows.


Nice to have

• You have built LLM agents, automations, prompt chains, or AI workflows for work, side projects, or your own productivity.

• You have experience with technical specs, structured outputs, JSON/API-shaped data, evals, QA, workflow builders, or lightweight scripting.
• You have worked at an early-stage startup, started your own company/project, or operated in a high-ownership environment with lightweight structure and processes.
• You have experience in healthcare, healthcare operations, reimbursement, prior authorizations, appeals, RCM, insurance, or clinical documentation.

Example projects

These are projects you could work on, but specific work will vary based on your strengths and preferences.

• Platformizing our customer-specific AI workflows. Help turn customer-specific agentic workflows we've built for early customers into a more centralized system of reusable prompts, schemas, evals, document logic, and implementation patterns that can be applied across future customers.
• Making our agents less "over-cautious." When we first built these agents last year, LLMs didn't have enough judgment to distinguish which details are critical in a patient case vs. which details are relevant-but-not-critical. But in the last couple months, LLM "judgment" has improved, and we want to experiment with giving that judgment to our agents without sacrificing enterprise-grade accuracy.
• Expanding our Voice AI agent. We've successfully deployed Voice AI agents in a customer pilot, but we want to expand, You can help define where voice agents can reliably handle outcome monitoring, payer/provider follow-up, status checks, and other operational workflows.
• Exploring computer-use agents and UI automation. Test where browser-based or UI-path agents can safely navigate the internet and existing software, complete repetitive tasks, retrieve case information, and extend Paxos's automation surface area beyond document generation.
• Building the Paxos AI workflow playbook. Create the reusable internal system for how we document workflows, write agent instructions, define edge cases, structure outputs, review failures, and turn customer-specific implementations into repeatable product patterns.

About the team

More on us here.

• Haley King , CEO & Co-Founder: Haley spent seven years at Medtronic across engineering, product strategy, and FDA/regulatory work. She holds an MBA from Stanford.
• Alex Lacey , CPO & Co-Founder: Alex was a product manager at Meta and Microsoft, has been a serial entrepreneur since high school, and holds an MBA from Stanford.
• Malcolm Asher , Head of Ops & Co-Founder: Malcolm has spent years working on patient access, including founding a global youth-led healthcare nonprofit. He's a graduate of Stanford University.
• Ben Setel , Founding/Lead Engineer: Ben leads Paxos's engineering efforts and brings a mix of software engineering, AI, startup, legal, and compliance experience. Before Paxos, he founded a machine learning startup.
• Vikram Karuna , Founding Engineer: Vikram has 15+ years of software engineering experience, including at Microsoft, and he previously founded an AI startup. He studied at the University of Waterloo.

Working style: You can expect a fast-paced, intense, dynamic, and exciting working environment where everyone has strategic ownership. We are a small team where everyone works directly with the founders. We take our work seriously, but we don't take ourselves seriously, and we are building a culture that prioritizes constant learning and feedback. We also play online games like Geoguessr at the end of every standup.

Compensation

• $110,000-$175,000 USD base salary, calibrated to candidate level
• Very generous equity package, reflecting our belief that the high ownership given to our early team should come with meaningful upside
• We offer health, dental, and vision insurance, plus standard paid time off per early-stage startup norms