Hi ! I’m Goutham, an ML Engineer based in the SF Bay Area, currently at Gen Digital (the company behind Norton, Avast, and LifeLock).
My work sits at the intersection of machine learning, security, and large-scale systems. Day to day, that means building AI-powered services, fraud detection pipelines, identity protection systems, and increasingly agentic AI infrastructure for a platform that protects millions of users worldwide.
Lately like pretty much everyone in AI right now, I’ve been deep in the agentic AI stack: multi-agent orchestration, deepresearch agents, agent trust boundaries, and the hard problem of evaluation at the harness level.
What I write about
This blog is where I think out loud about the things I’m building and learning. I find that writing clarifies my thinking, and I hope it’s useful to others navigating similar challenges. Expect posts on:
- Agentic AI - orchestration patterns, MCP, sub-agents, harness engineering
- Production ML - evaluation, fraud detection, feature stores, inference at scale
- Security x AI - agent trust, prompt injection, policy enforcement, identity for non-human entities
- Side projects - architecture decisions, mistakes made, lessons from shipping
Background
I’ve spent 7+ years across ML engineering, NLP, and applied AI. Earlier in my career I worked on neural network accelerators on FPGAs. On the applied side, I’ve used ML to contribute to social good: building models to help match guide dogs with their humans at Guiding Eyes for the Blind, and supporting data collection and analysis work for the World Bank. I also built a computer vision system that estimates nutritional content from food images, one of those projects that starts as a fun experiment and teaches you more than you expect about real-world data messiness.
Let’s connect
I’m always interested in conversations about agentic AI, production ML systems, or anything at the frontier of security and AI. If you’re working on something in this space : research, a product, or just have a question, feel free to reach out.