A short
introduction.

I'm Harish, an engineer who lives at the intersection of mathematics, systems, and machine learning. Right now, I'm at Verily building ML infrastructure for agent orchestration workflows and retrieval-augmented generation systems. My work focuses on coordinating multi-step reasoning across agents and enabling large language models to access the right medical context.

Before that, I spent a couple years at Apple building distributed systems for test execution across their device farm. My work ranged from designing demand-prediction algorithms to building a workflow engine (from scratch in Swift) to scaling infrastructure that supported roughly a few million tests a day.

I studied Mathematics and Computer Science at Duke University, where I got deep into abstract algebra, topology, and real analysis while simultaneously falling in love with machine learning. My research focused on training population-based reinforcement learning algorithms given multi-modal human feedback.

These days I write about that intersection in the notebook, and tinker with small projects on the side. I also keep a quiet now page, a snapshot of where my attention is.

Reach me at hyerra@alumni.duke.edu, or find me on GitHub, LinkedIn, and X.