Writing
Thoughts & notes.
Occasional writing on AI engineering, product development, and the realities of building software end to end.
What I learned shipping 10 products solo
Three years, ten products, one person. What actually matters when you build end to end alone, and what doesn't.
Read →Building voice AI that doesn't sound like a robot
The difference between a voice demo and a production voice agent is about 200ms and a barge-in system. Here's what it actually takes.
Read →How to know if your LLM is actually working
LLMs in production degrade silently. Here's the eval infrastructure you need before you ship: regression tests, output scoring, and cost monitoring.
Read →7 layers to verify an email address
Building email-probe: from RFC 5322 syntax to live SMTP handshake. A deep dive into the verification pipeline that cut bounce rates by 95%.
Read →Don't let your LLM do math
LLMs are terrible at arithmetic. Here's how I built a financial data platform that uses AI for everything except the numbers, and why that matters.
Read →Building agents that govern their own output
The most important feature of RankOps isn't the AI that writes SEO metadata. It's the system that reverts bad changes. Autonomous agents need self-governance.
Read →