From Cisco to FarmChem

TLDR: Built AI agent pipeline for proactive cybersecurity mitigations at Cisco. Next up: industrial automation software at FarmChem + AI infra on my own time.

That's a wrap on my summer as an AI Machine Learning Engineer Intern at Cisco! 🎉

This summer, I worked on agentic AI systems for proactive cybersecurity defenses under my manager, Vinny Parla, who scoped the project with me and kept me on track with weekly check-ins. Working closely with Ganesh Sundaram, Kai Cheung, and the HYPERAI team, I built a pipeline that used AI agents to generate and validate policy mitigations, turning a slow, manual process into something fast and reliable. I worked across the full pipeline, from retrieval and generation to validation and learning about the Kubernetes/AWS infrastructure that ran it all at production level.

This was my first real lesson that the validation layer and guardrails matter as much as the generation itself, something Kai Cheung drove home for me again and again. I also got to present my work twice to Nancy Cam-Winget, which pushed me to get sharper at communicating technical work to a broader audience. I learned more about security innovations in 12 weeks than I expected to learn in a year.

Huge thanks to Rajvardhan Deshmukh for helping me get set up and find my footing at the San Jose office from day one, and to Lichi Li for being great to work and talk about cars with. Thanks also to Craig Connors for building such a great team to be part of, and to Carlos Diaz for the career advice and for having my back through my presentations. Thanks as well to my fellow interns on the HyperAI team, Shakthi Bala and Kristen North, for all the shared meetings with the team and great conversations. It was also great getting to know Shaobo Gu and John Alexander Carr on the team from dinners to outings.

Next up: I'm headed to a part-time software engineering internship at FarmChem during the semester, where I'll be writing industrial automation software: code that controls and coordinates physical equipment and processes. Different problem space, same excitement. I'll also be spending my free time going deeper on AI infra and model training.