My background

I started as an electrical engineer working on data analysis and signal processing—building intuition around signals, noise, and randomness from day one. I worked across wireless industries, from communication satellites to high-end commercial WiFi routers, constantly building hardware and software automation to characterize and test end-to-end systems, using software.

That's when I fell in love with software. I realized it was the glue that bound complex systems together. Software led me to data science, where I spent four years productionizing forecasting models before moving into marketing to build automated messaging platform.

In marketing, I learned about A/B testing frameworks and dove deep into experimentation science. That's where it all clicked: everything I'd been doing—from signal processing to forecasting to experiments—was really just about separating signal (ie information) from noise (ie uncertainty).

What I'm interested in

• Turning complex statistical concepts into practical engineering tools
• Agentic orchestration and eval design
• Experimentation design and causal inference
• Building smart large-scale distributed systems
• Probabilistic programming

Why I started this notebook

Engineers should own the end-to-end experience of the systems they ship — which means measuring what the user actually sees and working backward through every layer that can move that signal. You can't improve what you can't measure, and you can't measure well without statistical intuition about noise, sampling, and failure at the edges.

My goal here is to write about data, models, and decisions the way engineers actually think about them — through real systems, monitoring, experiments, and things that break — not textbook jargon.

A few things I keep coming back to:

• Most ML problems are data problems.
• Aggregate metrics hide real behavior.
• Systems fail at the edges, not averages.
• Better inputs beat bigger models.