I’ve been building software for thirty years. I’ve been around long enough to remember when “cloud” was a sales pitch and “AI” meant a decision tree. I’ve watched several technology cycles get declared revolutionary and then quietly absorbed into the infrastructure.
Right now I’m Director of Language Engineering at Talos Title, building multi-LLM orchestration systems for legal document processing. Before that, eleven years at Verizon, a FinTech startup called Setscale where six of us built the thing from zero, and a bunch of other stops along the way. Three decades of making technical decisions that had real consequences, and learning which ideas hold up under production load and which ones collapse on contact with reality.
I’ve been on LinkedIn for a while. It works for staying visible and occasionally saying something useful. But LinkedIn has a gravity problem: the platform rewards short, punchy, optimized-for-engagement content. Write something nuanced and the algorithm buries it. Write something with actual depth and you lose half your audience before the second paragraph.
I wanted somewhere to write longer pieces. Pieces that don’t have to survive the scroll. Pieces that can take a position, develop an argument, and exist somewhere beyond a 48-hour feed window.
That’s why Substack. Not because I have a brand to build. Not because I’m positioning myself as a thought leader — a phrase that has lost all meaning and should probably be retired. Just because I have things I actually think, and I’d rather write them somewhere they can breathe.
Here’s what you can expect: one piece every two weeks. The topics will move around, but they’ll stay in a few lanes.
LLM orchestration architecture — the real engineering decisions, not the whitepaper version. What it actually looks like to build systems where multiple models hand off to each other, where failures need to be recoverable, and where the output has legal weight.
Building AI in regulated industries. Title insurance isn’t a sandbox. When a system touches a real estate closing, the tolerance for hallucination is approximately zero.
Career reflections from thirty-plus years in the industry. What I’ve gotten wrong. What I’ve changed my mind about. What I think is genuinely different about this moment versus the last several times someone declared software was being fundamentally transformed.
If you build AI systems, lead engineering teams, or are trying to figure out what it looks like to deploy this technology in a high-stakes environment where mistakes have consequences — this is written for you.
If you’re here for inspiration content, you’re going to be disappointed.
I’ll see you in two weeks.