From the EP lab to the claim.
I supported 150+ cardiac ablation cases at the bedside and spent five years in med-device systems and quality engineering. Now I build AI-assisted tools for the documentation, coding, and denial work that decides whether that care gets represented and paid correctly.
Working tools
Each one runs in the browser on synthetic data, shows the evidence behind every suggestion, and says plainly what it does not do.
Site-Tailored CDI Optimizer
Finds CC/MCC conditions with clinical evidence but missing or vague documentation, using de-identified ICU data from two very different health systems.
- Problem
- One national rule set over-queries some hospitals and misses cases at others. Physicians at each site agree with queries at very different rates.
- Does
- Applies each site's own policy, learns its physicians' agreement patterns (model AUC drops from about 0.7 to about 0.5 across sites), ranks queries by DRG and severity impact, and drafts compliant queries with a grounding check on the AI layer.
EP Ablation Coding Auditor
Reads a cardiac ablation note and returns CPT and ICD-10-CM suggestions, each tied to the sentence that supports it.
- Problem
- Since the 2022 bundling changes, ablation add-ons are a common source of both missed revenue and audit risk, and the difference is usually one missing sentence.
- Does
- Applies 93653/93654/93656 bundling, flags unsupported add-ons, scores documentation, and drafts non-leading CDI queries.
Denial Triage & Appeal Builder
Turns a denial worklist into a ranked queue with the next action, the owner, the deadline, and a drafted letter.
- Problem
- Denial teams work queues by age, not by recoverable dollars or deadline, and the root cause rarely makes it back upstream.
- Does
- Classifies 35+ adjustment reason codes, computes appeal windows, ranks by expected recovery, and recognizes EP bundling denials that should not be appealed.
Upnea Wireless ST
Wearable screening and diagnosis concept for obstructive sleep apnea. I led engineering and also drove needs finding, market analysis, and pricing, including the reimbursement strategy.
- Context
- OSA carries an estimated $144B in indirect costs, and most cases go undiagnosed because in-lab sleep studies are a bottleneck.
How I build with AI
AI makes building fast. In healthcare, checking the result matters more.
Start from the user's chair
Write down the decision a coder, CDI specialist, or denials analyst makes, and what they need to see to trust a suggestion.
Pair with Claude
Draft the rules, the UI, and the edge-case test notes with an LLM so a working version exists in hours, not weeks.
Check primary sources
Check every rule against CPT Assistant, CMS manuals, and payer policy. If I can't cite it, it ships as "needs review."
Break it on purpose
Negations ("ICE was not used"), redo cases, and missing-sequence notes. The same habits I used writing design verification at Abbott and Boston Scientific.
Experience
Clinical field work, quality, systems engineering, and R&D across three cardiovascular device companies.
Affera Mapping Specialist
- Supported 150+ cardiac ablation cases in the EP lab, averaging 12–15 cases a week in high-volume accounts.
- Ran the 3D mapping system for physicians during PVI, CTI, and substrate cases; helped a new physician adopt the platform, who went on to do a few dozen cases.
Design Quality Engineer II
- Design controls, risk management, and design verification documentation for medical devices.
Systems Engineer II · R&D Engineer
- Systems engineering for the CardioMEMS Patient Electronics System and R&D on the PA pressure sensor and delivery catheter.
BS Biomedical Engineering · Master of Biomedical Innovation & Development
- Earlier roles at CryoLife, Motus Nova, and W. L. Gore & Associates.
Let's talk.
Solutions, product, and clinical-ops roles in health tech, especially revenue cycle AI.