Aaron Murdock

Aaron Murdock

Product Leader, AI + Digital Health

I'm Aaron, a product leader who builds consumer digital health products that change behavior and move clinical outcomes. Right now I'm most interested in what AI can actually do for patients, which is why I build the prototypes instead of just writing specs about them.

Portrait of Aaron Murdock

Some fun projects I have worked on

02

Gestational Diabetes App

Try the prototype ↗

Product walkthrough

{{ slide.alt }}
{{ slide.boxLabel }}

{{ activeCaption }}

Problem

Gestational diabetes is managed almost entirely by the patient, and the instruction is essentially “keep your numbers down.” But the feedback loop is broken. She eats, and the consequence shows up two hours later as a number with no label on it. Without knowing which meal caused which spike, she’s running the same experiment over and over without ever seeing the result.

Solution

An app powered by a continuous glucose monitoring device that closes that loop automatically. It correlates every logged meal with the glucose readings inside the American Diabetes Association's 2-hour post-meal window, so each meal comes back with its own answer instead of a number floating free. It flags what’s out of range, and when glucose climbs past 140 with no meal on record, it asks her what she ate rather than letting the data point go unexplained. Everything rolls up into a report she can hand to her care team.

Role

Personal passion project. Concept, product design, and the working prototype.

03

Group Therapy Platform

Try the prototype ↗

Product walkthrough

{{ slide.alt }}
{{ slide.boxLabel }}

{{ activeCaption }}

Problem

For gambling addiction, group CBT performs at least as well as individual therapy. It’s also cheaper to deliver and expands a clinician’s reach. Yet no specialized telehealth network offers it as a real, insurance-billed modality, so patients get 1:1 care by default. The barrier isn’t clinical, it’s operational: group sessions are harder to enroll, harder to keep attended, and harder to bill cleanly than a one-on-one appointment.

Solution

A platform that removes each of those operational barriers. Therapists refer patients into group care and patients self-enroll through a single flow that handles consent and insurance together. Sessions run as HIPAA-compliant multi-party video with roster management, automated attendance tracking, privacy controls, and safety escalation built in. Every session generates a claim-ready record, so the modality that already works clinically becomes one a network can actually run and bill.

Role

Case study for a Founding PM interview. Owned the market framing, the phased rollout recommendation, and the MVP scope. Presented to the CEO, VP of Clinical, and engineering.

04

Tennis Coach App

Try the prototype ↗

Product walkthrough

{{ slide.alt }}
{{ slide.boxLabel }}

{{ activeCaption }}

Problem

Getting to 4.0 rewards fixing obvious technical flaws. Getting past it doesn’t. At that level the gap is strategic and physical, not mechanical, and it’s specific to how you actually lose points. But coaching content is generic by nature, and a weekly lesson can’t see the match you played Saturday. So players drill what they’re already good at and plateau for years.

Solution

An app with an AI chat-based coach that starts where the plateau actually lives: your last match. It debriefs how the match went, then turns that into two things — what to do differently strategically in your next match, and the specific drills and conditioning to work on in the days between. The plan changes as your game does, because it’s built from your matches instead of a generic curriculum.

Role

Personal project, solo build.

01

Pregnancy Management Platform

Try the prototype ↗

Product walkthrough

{{ slide.alt }}
{{ slide.boxLabel }}

{{ activeCaption }}

Problem

The U.S. pregnancy care model has barely changed since the 1930s: a fixed schedule of in-person visits, with long stretches of no visibility in between. Meanwhile providers are absorbing more administrative work, more payer red tape, and more patients, which leaves less actual time with each one. The result is a care model built for a world that no longer exists, run by teams who have less time than ever to work it.

Solution

A platform that keeps a care team connected to pregnant patients between visits. Patients receive integrated remote monitoring devices, and their readings flow continuously to coordinators instead of surfacing only at the next appointment. A generative AI chat agent handles routine administrative tasks to help you manage your patients better within clinical guardrails, so the care team's limited time goes to the patients who need it. A CPT coding engine turns that monitored care into billable encounters without adding administrative load.

Role

Director of Product at Lōvu. 0-to-1 work across care coordination, patient engagement, and the CPT coding engine. Built this prototype as a redesign of the core platform to inform product development during the company's growth stage.

Hi, I’m Aaron.

I like product problems that are hard and actually matter. Digital health is where I’ve spent my career, building things that change behavior and move clinical outcomes, not just engagement charts.

Sales, then customer success, then Director of Product in under three years. The through-line is that I’ve never wanted to be the person who writes the spec and walks away. Product and engineering work best as one unit.

Along the way I’ve built clinical assessment tools and owned EHR integrations (HL7/FHIR across Epic and athenahealth) which is where I learned how much of a health product lives in the data plumbing, consent, and billing paths nobody demos.

AI is the part I’m most fired up about right now. Everything on this site I built myself, because I’d rather prototype the idea than argue about it in a doc.

Outside of work I’m on a tennis court most weeks, a dedicated father (to my animal children), learning guitar badly, playing video games, hiking, and taking my Jeep somewhere it doesn’t belong. Also planning a wedding, which turns out to be a roadmap with a very fixed deadline.

Tabby cat lounging on a wall-mounted cat shelf
Aaron and his fiancée on a hill country road
Grey longhaired cat stretched out on a chair
Aaron and his fiancée at sunset by the ocean
Blue-eyed cat sitting in a sunny window

Based in Austin, TX. Always up for a conversation about AI in health — say hello.

How I operate

I don't wing product work. Every problem I take on runs through the same three loops: find what's real and worth solving, build and ship it with the team, and diagnose what actually happened once it's live. These aren't slides I pull out for interviews, they're how I actually operate, on every team I've been on.

Loop 01

Discover, Strategize & Prioritize

How I frame problems, gather evidence, and decide what earns a place on the roadmap

When I use it

I'm identifying opportunities, framing problems, and deciding what's worth building next.

The steps

GroundCollectFilterScoreSequenceCommit

  1. 1.1

    Ground

    Understand goals, constraints, and context before chasing solutions.

    ToolsOKRs & company strategy · Leadership & clinical stakeholder interviews · Competitive & market scan · Prior roadmap & retro notes

  2. 1.2

    Collect

    Gather quantitative and qualitative evidence, then triangulate the two.

    ToolsMixpanel / PostHog / Amplitude (usage & funnels) · NPS & CSAT trends · Support ticket themes · User interviews & app store reviews · Sales & CS feedback loops

  3. 1.3

    Filter

    Narrow the noise down to the problems actually worth solving.

    ToolsAffinity mapping · Opportunity solution trees · Jobs-to-be-done framing · Quant + qual triangulation

  4. 1.4

    Score

    Rank the surviving problems and ideas by impact versus effort.

    ToolsRICE / ICE scoring · Value vs. effort matrix · Weighted scoring vs. North Star metric & OKRs

  5. 1.5

    Sequence

    Order the roadmap with design and engineering, who ground it in feasible reality.

    ToolsSequencing workshops with eng & design leads · Technical feasibility spikes · Dependency mapping · Capacity planning vs. sprint velocity

  6. 1.6

    Commit

    Align partners, name the bet, and write down the risk.

    ToolsOne-pagers / PRDs · RAID log (risks, assumptions, issues, dependencies) · Roadmap review with execs & stakeholders · Go / no-go criteria

In practice · Lōvu

Engagement was slipping at Lōvu. Collect showed why: our Bluetooth RPM devices had a lag between a mom taking a reading and it hitting our system. That killed trust with patients and partner clinics, and it slowed how fast we could escalate urgent readings. Filter and Score turned that into a clear bet. I moved us off all 3 Bluetooth devices onto 4 new cellular ones. No pairing, ready out of the box, no need to be near a phone to transmit.

Loop 02

Execute & Ship

How I turn a committed bet into something shipped, instrumented, and real

When I use it

Strategy is set, the bet is named, and it's time to deliver.

The steps

ScopeAlignBuildValidateShip

  1. 2.1

    Scope

    Align on the problem and why it's worth solving before anything else. Then the goal, user segment, and what's explicitly in and out.

    ToolsProblem statement & why-now case · PRD / spec doc (Notion or Confluence) · Success metrics defined upfront · User stories mapped to JTBD

  2. 2.2

    Align

    Align on solution options, dependencies, tradeoffs, and risk with design, engineering, and stakeholders. That includes the assumptions, success metrics, and guardrail metrics that decide whether we iterate, roll back, or kill.

    ToolsDesign reviews in Figma · Engineering feasibility & architecture review · Dependency mapping with eng leads · Claude Code / Claude for Prototyping · Assumptions, risks & guardrail metrics (iterate / rollback / kill)

  3. 2.3

    Build

    Run the team's cadence and clear blockers fast.

    ToolsStandups & sprint planning · Backlog groomed in Jira / Linear · Daily pairing with eng & design

  4. 2.4

    Validate

    Test and UAT against requirements before anything ships.

    ToolsUAT with clinical & compliance reviewers · Usability testing · HIPAA / SOC 2 checks where relevant · Beta / TestFlight rollout

  5. 2.5

    Ship

    Release, instrument, then learn. Feed it back into discovery.

    ToolsFeature flags (LaunchDarkly) · Mixpanel / Amplitude event instrumentation · Post-launch monitoring dashboards · Retro & learnings loop

In practice · Everlywell

Building the AI IVR for a enterprise Payer Group, Align is where we locked the real numbers. 95% accuracy on the agent's talk track. 100% on verifying HIPAA identifiers (name, DOB, zip) with every single patient before the call moved forward. No exceptions. Guardrails were just as concrete: roll back to the standard IVR the second HIPAA verification failed, kill the program if kit return rates dropped, iterate if talk track accuracy fell below 95%.

Loop 03

Diagnose

How I find what changed, why, and what to do about it

When I use it

A metric moves unexpectedly or something just feels off.

The steps

ConfirmSegmentIsolateTestDecide

  1. 3.1

    Confirm

    Is the signal real? Pin down the metric, size, timing, and who's affected.

    ToolsMixpanel / Amplitude / PostHog dashboards · Data QA (check for tracking bugs) · Statistical significance check

  2. 3.2

    Segment

    Split by cohort, platform, and tenure to see where the change actually lives.

    ToolsCohort analysis (new vs. returning, tenure) · Platform / OS / app-version breakdown · Pregnancy-stage / clinical segmentation

  3. 3.3

    Isolate

    Rule out confounds and artifacts, then form a hypothesis.

    ToolsRelease / deploy timeline overlay · A/B test & feature-flag history · Support ticket & app store review scan

  4. 3.4

    Test

    Weigh the signals and separate what's real from noise.

    ToolsFunnel / event-level deep dive · Peer review with data & engineering · Confidence interval & sample-size check

  5. 3.5

    Decide

    Roll back or iterate against the guardrails we already agreed on. Then write down what we learned.

    ToolsGo / no-go with eng & stakeholders · Rollback via feature flag · Postmortem / incident writeup · Findings fed back into Discover

In practice · Lōvu

Migrating our RPM devices from Bluetooth to cellular tanked compliance almost overnight, about 30%. Segment and Isolate traced it to one cause. Patients had no idea why their device changed. Not a technical failure, a comms failure. Decide meant a targeted outreach campaign to about 5K patients. Compliance came back 30% above baseline, and onboarding got faster too.

Resume & contact

Open to product roles in AI, consumer products, and digital health products. The fastest way to reach me is email.

Resume & contact

Aaron Murdock — Resume

PDF · 1 page

↓ Download resume

Email · fastest way to reach me

aaronmurdock3@gmail.com

Certifications

  • Anthropic logo

    AI Fluency for Builders

    Anthropic

    Issued Jul 2026
  • Anthropic logo

    AI Fluency Framework and Foundations

    Anthropic

    Issued Jul 2026
  • Scrum Alliance logo

    Certified Scrum Product Owner (CSPO)

    Scrum Alliance

    Issued Jul 2024
  • Databricks logo

    Generative AI Fundamentals

    Databricks

    Issued Nov 2024
{{ zoomImg }}