Past Talks

BOSTON SECURITY MEETUP: “Rogue Agents and Honeypots” - APR’26

Designed for cybersecurity practitioners, this talk covered factors that drive AI agent security incidents, a flag for impact beyond an agent’s expected blast radius (demonstrated through experiments I and others have led). I shared experiences from monitoring for rogue agents in the wild. The call to action is to strongly govern, guard, and monitor all agents in our purview, what they can access, and what actions they can take.

“Kids don't care, they just try things” - Sept’25

I presented this lightning talk at a joint AI Safety Awareness Project (AISAP) x BlueDot Impact event to foster awareness of randomness in AI agents, the infinite options that can be output at any word, and the implications of secondary tendencies from LLMs. Using real-life examples (hallucinations and high-risk commands), this talk encouraged cautious and defensive methods for all agentic interactions (vs a risk-taking “YOLO” approach).

ZENITY/AI agent security summit: genetic adversarial coding- OCT’25

This talk raised awareness that self-improving agents can let attackers and defenders fine tune and maximize their objectives at inference time, for instance in secure coding (to introduce vs detect vulnerabilities).

Boston AI security meetup - JAN’25 to now

I’ve organized Boston meetups at various locations (in partnership with AI Safety Awareness Project, BlueDot, Harvard’s AISST and other local groups). This included situational awareness, technical deep dives, and most importantly lively discussions to exchange perspectives.

AKAMAI ACADEMY: BOTNETS KEYNOTE - MAY’18

I delivered the “botnets” keynote for Akamai’s annual internal conference for Security Services (Americas Region). In this talk, focused on situational awareness for technical teams interacting with customers, my co-presenter and I covered the Mirai botnet and threat actors’ Tactics, Techniques and Procedures for conducting (primarily) Distributed Denial of Service attacks. Mirai is soon turning 10 years old and is still a major active internet threat.