Leadership - Physician. Product and technology leader.

I set product and technology strategy, own the roadmap, and lead the cross-functional teams that deliver it — with a clinician’s sense of where healthcare workflows break, and what it takes for AI to be safe inside them.

William Burke, M.D.

Partner, Flanders Forest

Harvard Neurosurgery · NYU Langone · Omada Health · Flanders Forest

How I lead - Find the problem worth solving. Build the least that solves it. Let the evidence decide what’s next.

I’ve led engineering, design, data and product teams of 2 to 25 people, working alongside the clinical operations teams they serve.

Strategy and roadmaps. Ruthless prioritization, starting from the root problem. Before anything is built, the team names the problem actually worth solving — and that defines the smallest product that solves it. The roadmap follows return on investment: we keep investing in an area while users confirm it is working, and move the team when a higher-value problem is waiting. See the EMR integration strategy that deliberately shipped only step one.
Product operating model. A repeatable loop: investigate and understand the top problem, build an MVP that addresses it, test it with real users to confirm it does, then decide — explicitly — whether to keep building in that area or move to one with a higher return. Every cycle ends in a decision made on evidence, so teams know what they are building, why, and when it is done.
Cross-functional teams. Engineering, design, data and product working as one team with clinical operations, not as a hand-off chain. Clinicians help define the problem and judge whether the solution works; engineers hear the problem first-hand rather than through a ticket. See the clinical report review gate.
Developing people. I’ve mentored mid-level engineers into strong individual contributors and on into product-minded tech leads — from taking the ticket they’re handed to asking the questions that uncover the real problem, and judging whether a proposed solution actually meets it. I’ve mentored product managers to think across the whole product ecosystem, so what they build integrates with the rest of the product and is designed to scale with the company.

Clinical AI - Clinical judgment, built into the system.

Medical training is why I start with the workflow rather than the model, and why I treat safety as something the architecture guarantees rather than something a policy asks for.

01
Clinical training
Medicine teaches where care actually happens — and where a wrong answer costs a patient, not a sprint.
02
The real workflow
I start from how clinicians and care teams actually work, and find the step where the workflow breaks.
03
AI that fits it
The model takes the reading and drafting off people’s plates; the clinical judgment stays with the clinician.
04
Safe by construction
The safety guarantees live in the system’s architecture, so they hold even when someone makes a mistake later.

Principles, and where they were applied

Automate the read, never the judgment. Scope AI by what a wrong answer costs in each direction. The model reads and triages; a person keeps the call. Read the case study →
Unsafe output should be unreachable, not discouraged. An AI-drafted report can’t reach a patient before a physician signs it — because the data makes it impossible, not because a rule says so. Read the case study →
The boundary belongs in the system, not the prompt. A clinical copilot that only sees the patient in front of the clinician, enforced where the data is stored rather than by instructions to the model. Read the case study →
When the rules are unclear, refuse to guess. Where a decision carries regulatory weight, the system blocks rather than emitting an answer it can’t stand behind. Read the case study →

Selected work - The challenge, the decisions, and what changed.

Client details are kept confidential, so each case study describes the impact rather than the figures behind it.

Metabolic Health Platform · Fractional CTO

The Data Model That Makes an Unreviewed AI Report Unreachable

A clinical assessment pipeline where an LLM drafts a report, deterministic logic owns every medically consequential decision, and the data model — not a status flag — makes it structurally impossible for a patient to see a draft before a physician signs off.

Regulated AI & Clinical Safety Architecture

Otto Surgical AI

The EMR Integration Strategy That Deliberately Shipped Only Step One

A four-rung EMR write-back ladder, each rung independently shippable, with an explicit exit trigger for climbing to the next — sidestepping the integration trap that quietly burns runway at most early health-tech companies.

Clinical Product Strategy

Primum

Automating the Read, Never the Judgment

An AI triage layer for adverse reaction reporting, scoped by a single question — what does a wrong answer cost in each direction — so the model does the reading and a human keeps the call.

AI Operations

Otto Surgical AI

Resolving the Riskiest Unknown Before Writing a Line of Product Code

A focused four-day spike that tested whether consumer AirPods could deliver clean, surgeon-isolated audio in a loud operating room — before any engineering was committed to the hands-free capture path it would have to support.

Technical Risk De-risking

Tell us about your project

Get in Touch

Our offices

Portland
220 NW 8th Ave
97209, Portland, Oregon