About
Hey, I’m Chris! 👋
I’m a software engineer at Microsoft. Most of my work is getting AI systems to run in places they’re not supposed to work: air-gapped government networks, on-prem servers with no internet, sovereign clouds where you can’t call anyone else’s API. In practice that means chat apps, retrieval over private data, and agents that use tools while running on models that don’t always cooperate. It’s messy and constrained and I genuinely love it.
The thing I can’t stop thinking about, though, is what’s going on inside these models. I’ve been drawn to interpretability since my internship at Novetta in 2020, where I trained GPT-2 to explain its own predictions. These days I run small experiments when I get curious about something, like why reasoning models fight so hard to fix errors in their chain of thought, or whether telling a model to act like a goblin changes how it codes. I write them up here because I think more people should poke at these things and share what they find.
I also can’t stop building things for myself. It’s a problem. I wrote up a philosophy for designing CLIs so AI agents can drive them, and I built a pixel-art office for managing coding agents because four terminal tabs was driving me insane. There’s also an AI coworker that sends me a morning brief, watches my Teams, and runs 30+ daily workflows. At some point the side projects became the portfolio, which wasn’t the plan, but I’m not complaining.
Before Microsoft I taught data science at Forge for two years. There’s something about watching a student’s face when a concept clicks, and I still chase that feeling. I built a 1,200-person early-career engineering community at Microsoft for the same reason. Helping people find their footing early is something I’ll probably never stop doing.
The personal stuff
I grew up in Tampa in a loud Italian-American family. Oldest of four, which means I’ve been project-managing people since before I knew what that was. I live in Seattle now with a kitchen that’s seen some ambitious dinner parties.
I plan restaurant reservations three weeks out, and I show up to a new city with a ranked list of coffee shops. More than once I’ve built a custom cocktail menu for a Tuesday night. Most of my best nights have started in that kitchen with way too much food.
I’m particularly interested in retrieval systems, agentic infrastructure, and interpretability research. Always happy to talk, so find me on LinkedIn.