Devesh S.

The Question I Can't Stop Asking

"We know exactly what our EHR does when it fails. We have no idea what our AI does when it fails — because it hasn't failed loudly enough yet." A hospital CIO, in conversation

That sentence is the reason this site exists.

I didn't start in healthcare. I started in telecom, staring at a harder version of the same problem — four of the largest carriers in the US, all fighting networks that retrained themselves on new traffic and attack patterns faster than any human team could monitor. I architected the Cyber Security Intelligence Solution that gave AT&T, Verizon, Comcast, and T-Mobile a way to catch threats inside that moving target. It opened roughly $40M in new business. The number I actually care about is different: it proved that security bolted onto AI after deployment always loses to security built into the architecture from day one.

I call that principle Secure by Architecture, Not by Patch — it's the lens I now put on every AI system I touch, and the thesis of my PhD.

I finished a PhD in AI & Cybersecurity in 2025 because I kept running into a version of that CIO's sentence in every regulated industry I've worked in. Healthcare is where it's loudest. Health systems are deploying generative AI into clinical workflows faster than governance frameworks can keep up — HIPAA wasn't written for an autonomous agent making a clinical judgment call on PHI in three seconds. That gap is where I now spend most of my reading, writing, and advisory conversations.

Today I lead Cisco's Worldwide Cloud & AI Infrastructure Sales organization, working with CXOs and architecture teams to design AI infrastructure — Cisco UCS X-Series, NVIDIA integrations, Cisco's Secure AI Factory framework — that's governable from the first line of code, not retrofitted after an incident report.