About

The standard data science playbook is obsolete.

I realized this during my first year managing AI systems at Google.

Building a ML/Statistical model matters less than knowing how to nail the design and the buiding guardrails that keep the model from lying.

This site exists because the ground is shifting beneath us, and the industry is still pretending the there are massive productive gains just because execution is cheap.

What I Do

I am a Research Data Scientist at Google, specializing in the architecture, evaluation, and automated guardrails of next-generation AI systems. My day-to-day work focuses on building scalable evaluation infrastructure—leveraging LLM-as-a-judge (Autoraters), human-in-the-loop (HITL) frameworks, and agentic workflows to measure and optimize ambient intelligence and voice experiences.

Before tackling generative AI quality at Google, I spent years navigating data velocity and algorithmic complexity at scale within Twitter and Walmart. My focus has consistently been on turning volatile data environments into predictable product value—whether optimizing massive e-commerce and social graphs, or engineering the complex geospatial analytics that empowered a million households to transition to solar energy. I’ve authored two books (Geospatial Science Quick Start Guide and JavaScript for Developers by Example) and hold a patent in transportation graph analytics.

What I Believe

The title "data scientist" is being aggressively redefined. The future does not belong to those who merely write code or prompt models. It belongs to the architects who can orchestrate complex AI systems, design rigorous experimentation frameworks, and ruthlessly identify when an intelligent system is confidently wrong.

The historical walls between data science, product management, and software engineering are collapsing. This is an exceptional opportunity, not a threat. An engineer with statistical intuition can own a feature end-to-end; a product manager with experimental rigor can steer a model's direction. This newsletter is designed for the practitioners who want to lead that cross-functional convergence, not just watch it happen.

I do not write to give you generic tutorials or hype. I write to provide the hard, statistical truth about deploying production-grade AI.

Current Projects

I am currently writing my next book, The AI-First Data Scientist. This newsletter is where the underlying frameworks, architectural debates, and core methodologies are pressure-tested first.

If your goal is to gain true technical agency and operate as an AI architect—rather than chasing superficial AI vibes—you belong here.