I flew KC-10s in the Air Force. Not as a pilot — as a flight engineer. The job is systems management and crew coordination: you don't fly the plane, you keep everything that makes flight possible running at spec. Fuel, pressurization, hydraulics, electrical — and you're the one running the checklists, calling out the steps, making sure the right information gets to the right person at the right time. You're watching the thing behind the thing.

That habit of attention — watching for what's not visible on the surface — turned out to be the most useful thing I carried into a career in data analytics.

I've spent years as a Senior Data Analyst at a large SaaS enterprise. When ChatGPT launched, I was an early adopter. When AI coding tools showed up — Aider, Cline, Cursor — I was already using them before most people knew they existed. I was paying attention.

Then my company gave everyone Cursor, and I watched what happened. Smart people, cutting edge models, and the result? Mostly noise. The queries were wrong. The outputs needed hours of cleanup. The AI didn't know anything about our data, our systems, our work.

It wasn't a people problem — the talent was there. It was an infrastructure problem. Nothing in the environment was built to give the AI the context it needed to be useful. The model wasn't the gap. Context was the gap.

In signal processing, the carrier wave is the dominant signal — it has the most power, but it carries no information by itself. The information lives in the sidebands: the frequencies created when something meaningful modulates the carrier. The AI model is the carrier wave. The context from your data, your systems, your workflows — that's the sideband. And nobody was building it.

So I started building it. A Snowflake MCP so the AI could query actual company data. A Jira MCP so it could see what was actually on the roadmap. Email, calendar, Confluence, web extraction — each one a layer of context that made every other tool sharper. First build took days. Now I can ship a new MCP in an afternoon. The learning compounds the same way the tools do.

That's the mission behind Sideband Labs. The tools, the thinking, and the conviction that the gap between "AI can do this" and "AI just did this at work" is a context problem — not a model problem.