I spent seven years as a senior systems engineer at FreeWheel, a Comcast company, working inside server-side ad insertion (SSAI) for streaming video at national scale. Before that I worked on national video infrastructure at Comcast and Charter. Ad tech is where I learned how production systems really fail. Now I bring AI to the parts of that work that have always eaten engineers' time.
Where the Time Goes
In streaming ad operations, the hard problems rarely show up in a dashboard. A partner reports that ads aren't rendering on one device family. Fill rate drops in one market. A traffic segment starts looking like invalid traffic. Each one starts the same way: someone pulls logs from several systems, lines up timestamps, and works out which link in the chain broke.
I've led those investigations, coordinating engineering, product, and client teams to find the root cause. I built SSAI diagnostic tooling that cut resolution time and became a go-to resource for the wider engineering team. I also automated ad-inventory reporting covering render rates, IVT, and traffic segmentation, replacing hours of manual spreadsheet work with Python and pandas.
What I Build
- AI-assisted incident triage — agents that pull and correlate ad-server, SSAI, and player logs, then put together a first-pass timeline before an engineer opens a ticket
- Reporting automation — inventory, render-rate, and traffic-quality reports that build themselves, with a language model writing the "what changed and why" summary
- Diagnostic tooling — focused internal tools that answer the same questions every on-call engineer asks, so they don't have to be rebuilt during every incident
- MCP access to ad data — scoped, read-only Model Context Protocol servers over your reporting databases and log stores, so AI assistants can query real data instead of guessing
- Partner-ready communication — draft incident summaries written for business stakeholders, not just engineers
How I Approach It
Ad tech runs on money and trust, so I keep the AI parts conservative. Agents read and summarize. People decide and act. Access is read-only unless there's a clear reason otherwise. Every summary links back to the raw records it came from, so an engineer can check the work in seconds.
"The best incident tool I built at FreeWheel didn't do anything clever. It asked the questions every engineer asked first, automatically, and had the answers ready when they arrived."
That's the bar for AI here: it should take away the first hour of every investigation, without becoming one more system to debug.
Good Fits
- Ad operations and support teams buried in repeated investigations
- Publishers and platforms with manual inventory or partner reporting
- Engineering teams that want AI assistants with real, scoped access to ad and streaming data
- Anyone who needs a consultant who has run SSAI incidents, not just read about them
Ad Ops Taking Too Many Hours?
I take on consulting engagements and focused builds for ad-tech and streaming teams. Let's look at where your engineers' time goes.
Start the Conversation