Phil McCaffrey

Every U.S. hospital is federally required to publish its prices — and they do, in thousands of incompatible files no patient could ever read. ClearPrice turns that chaos into a simple answer: what will this cost me, here, with my insurance?

• Took healthcare price transparency from federal mandate to working product: a 50-state platform covering 7,110 hospitals and ~5.7 billion insurer-negotiated rates in a ~1.9 TB production database — conceived, architected, built, and operated solo in five months
• Solved the problem that makes the mandate useless in practice — every hospital publishes differently — with a format-agnostic ETL pipeline that auto-detects and streams 20+ file formats, from 5 GB JSON dumps to legacy Excel
• Scaled from 1 state to all 50 in three months on an event-driven AWS architecture (ECS Fargate Spot fleets, 100 per-state SQS queues, EventBridge) deliberately designed so a single engineer can operate a national data platform
• Made 5.7 billion rows conversational: a chat-first consumer app, advisor portal, and ops dashboard powered by 27+ specialized AI tools and 10 visual React components, with a tiered Claude model policy and prompt caching that holds cost-per-question flat
• Turned raw compliance data into something trustworthy: 4,457 payer-name variants consolidated to 1,702 canonical payers, every hospital graded A–F on data quality, and Medicare coverage intelligence spanning 6,839 Advantage plans
• Proved an engineering org of one can ship with enterprise rigor: 200 architecture decision records, 322 test files, multi-agent AI code review, and 7-minute push-to-live CI — an AI-augmented process engineered as deliberately as the product itself

Tech: Python, FastAPI, PostgreSQL/Aurora, React, TypeScript, AWS (ECS, SQS, EventBridge, S3), Anthropic Claude API

Companies from startups to Netflix hand this practice the problems they can't staff — one person owning each end to end: architect, builder, vendor gatekeeper, closer.

• Advised clients including Netflix, Lessen, AvAir, and Global Franchise Group on architecture, hiring, and scaling; delivered AI- and OCR-powered document processing platforms for legal and medical clients with mixed contractor and direct-hire teams

Order and inventory engineering for the biggest live events on earth — where launch day is non-negotiable.

• Led the distributed order and inventory systems behind the Olympics, F1, and the O2 Arena, then opened that inventory to third-party ticketing services for international venues and events
• Opened four new business lines for the platform — hospitality, travel, merchandising, and upgrades — architecting and launching domains powered by acquisitions and strategic partnerships
• Brought AI-augmented workflows into development and QA ahead of the industry curve — a discipline later scaled into ClearPrice's engineering org of one
• Drove organization-wide adoption of Domain-Driven Design — creating and owning the company's first core business domains, Orders and Inventory
• Established OKR frameworks and KPI dashboards for executive reporting, and served on the Scaled Agile advisory committee shaping how the technology organization ran SAFe at scale
• Built and scaled engineering teams across NA, EU, and APAC — fully remote and hybrid pods — while rebuilding the hiring pipeline for higher candidate quality and faster time-to-hire

Tech: Java, Python, Node.js, React, AWS, Kafka, Cassandra, GraphQL

• Led four independent cross-functional engineering business units — owning every customer-facing lead-generation platform, client portals, and the corporate web presence
• Set company-wide UI/UX standards and introduced BDD — moving the organization from instinct to evidence by aligning stakeholders, engineers, and QA around executable requirements
• Launched a software apprenticeship program to grow the company's own junior engineering talent

Tech: React, Java, AWS, Adobe Experience Manager, Adobe Analytics

• Directed mobile and CMS development, and found $1.5M in annual savings hiding in third-party contractor spend on data ingestion and reporting pipelines
• Rebuilt the technology investment and prioritization strategy across Best Western's tech organizations, grounding C-level decisions in hard data
• Consolidated fragmented analytics into timely, actionable business intelligence
• Cut new bugs 80% within 90 days by integrating BDD into QA

Tech: React, Java, AWS, Swift, Android, Cucumber

• Built and led the team behind Brand Registry — brand-protection services used by millions of people across hundreds of thousands of companies worldwide
• Developed SKU Central, collapsing what once took sellers 35 separate applications — 100+ data points and 70+ actions — into a single platform for contextual, high-confidence decisions
• Inherited a team three weeks from its first product launch and delivered — shipping to customers in time for the 2017 holiday sales cycle
• Stood up and scaled multiple distributed teams: a 12-person cross-functional engineering unit built from nothing in six months, and 13 engineers reorganized into three self-sufficient groups with dedicated scope and leaders
• Embedded BDD and automated QA so engineers, customers, and stakeholders shared one definition of done

Tech: React, Node.js, Java, AWS, Redis

• Designed and deployed real-time analytics processing 200M+ events a day across digital and in-person channels
• Led a team of up to 16 engineers integrating with 7 engineering teams across 4 international locations — plus marketing and product — with budget ownership and C-level planning responsibility
• Delivered "Merchant Insights," a location-intelligence BI platform built for internal planning and sales, and good enough that Groupon offered it directly to merchants and partners
• Shaped Groupon's company-wide adoption of Presto as a founding Advisory Board member

Tech: Java, PHP, Node.js, Redis, Kafka, Hadoop

• Invented and productionized an omnichannel analytics platform giving brick-and-mortar small businesses the user-event-level tracking that had been the privilege of e-commerce — millions of data points a day, five-nines availability, PCI-DSS Level 2 compliance
• Led technical due diligence through Groupon's acquisition, then drove the post-acquisition systems integration

Tech: PHP, Java, AWS

• Co-founded an engineering services firm and led its 30-person cross-functional team for eight years — software, managed IT, and a 200+ server datacenter — serving SMB and enterprise clients including Wells Fargo

An AI vocal coach that hears what a human coach hears — built solo in three and a half months.

• Built the full product: browser take-recording scored on pitch, timing, and dynamics with personalized AI coaching, powered by an ML pipeline for stem separation, lyric alignment, and pitch tracking on AWS Batch GPU Spot workers
• Engineered for near-zero idle cost — serverless, scale-to-zero AWS architecture — with prompt caching cutting LLM coaching cost per call ~85%
• Merged 621 PRs using the same multi-agent AI development workflow that powers ClearPrice

Tech: Python, TypeScript, AWS (Lambda, Batch, DynamoDB, Cognito), Terraform, Demucs, WhisperX, torchcrepe, Anthropic Claude API

Strategy sessions with AI deserve better than a scrolling chat transcript — so this renders them as living strategy documents.

• Built a browser-based AI consultation platform — FastAPI/WebSocket backend, build-free JavaScript client — rendering live advisory sessions as structured documents: risk matrices, decisions, and action items, driven by three pluggable consultant personas
• Deployed dual-mode, local-first and AWS, and used daily to drive product and consulting strategy decisions — ~14K lines, 107 automated tests

Tech: Python, FastAPI, WebSockets, JavaScript, DynamoDB, Terraform, Docker, Anthropic Claude API

Billable hours shouldn't fund paperwork — so this practice's paperwork runs itself.

• Automated the practice's time-and-billing back office: a scheduled Python pipeline over the Harvest and Microsoft Graph APIs drafts AI-written daily time entries, then generates and issues weekly invoices with expense reports — unattended

Tech: Python, Harvest API, Microsoft Graph, Anthropic Claude API, launchd