What Running 100 Miles Taught Me About Building AI Products
At mile 70 of the Vermont 100, everything hurts and nothing makes sense. That experience maps almost perfectly onto building and shipping AI products.
Hey, I'm Chris
From ski bum to AI product leader. I figure things out by building them.
Senior PM at LiveData, where I help build perioperative AI tools for 90+ hospitals. I have a habit of prototyping my own specs. Before tech I guided fly fishing trips, worked ski shops, did actuarial math on pension plans, and consulted on electric grid reliability. Turns out the least obvious path to AI product leadership was the most useful one.
Now
Tinkering with AI agents, shipping side projects, and writing about the things that surprise me. My latest build partner is my 8-year-old son. We're making a soccer trading card site together because finding card values is unreasonably difficult.
Writing
Thoughts on AI, building software, and the things in between.
What Building a Sprinter Van Taught Me About Product Management
I bought a used Sprinter 2500 and turned it into a family adventure vehicle. It was the best product education I've had since guiding fly fishing trips.
From Ski Slopes and Rivers to Product Strategy
How working in Colorado's outdoor industry taught me everything I needed to know about product leadership.
Projects
Things I've built, contributed to, or am currently tinkering with.
Enterprise AI Knowledge Engine
Built a RAG system that routes queries to department-specific context so Sales, CS, and Product each get answers grounded in their world, not a generic corpus. Designed two deployment tiers (developer-local and serverless) so both engineers and non-technical staff could use it. Includes an AI-powered feedback loop that triages corrections into the knowledge base automatically.
AI-Powered Site Health Dashboard
85+ hospital sites, each generating field reports as unstructured PDFs. Built an AI extraction layer that pulls structured data (issues, staffing gaps, compliance status) from documents field staff were already writing. No new process required. A health score algorithm surfaces which sites need attention without anyone reading 85 reports.
AI Sales ROI Calculator
A sales tool that generates facility-specific ROI projections with a key design principle: all math is deterministic Python. The AI interprets and narrates, but never touches a number. Three parallel AI agents handle analysis, storytelling, and objection prep. A conversational what-if agent lets reps adjust assumptions live in meetings.
Federal Surgical Data Pipeline
1.3M+ records from 10 federal sources (CMS, AHRQ, VA) unified into a single queryable schema. Built to put any hospital's surgical metrics in context: volumes, quality, costs, workforce. Idempotent, auditable, and designed so adding a new data source is one Python class, not an infrastructure project.
StrategyLive
Real-time AI sales coaching for Google Meet. Captures audio, transcribes locally via Whisper (nothing leaves your machine), and delivers live coaching suggestions through a Chrome extension overlay. An MVP exploring what happens when AI can listen to a sales call and help in the moment.
AI Content Repurposing Engine
Turn one transcript into 10+ pieces of on-brand content: LinkedIn posts, blog drafts, email sequences. The key decision was making brand voice a first-class input, not an afterthought. Generic AI content is the enemy of adoption. Built with tiered rate limiting and cost tracking from day one, not bolted on later.
Elsewhere
Other places you can find me on the internet.