Zeeshan Khan
Notes from building AI systems, and on the trust problems that define the next decade.
How do we know a system is doing what it claims to be doing? Most of what I write here comes from asking that question across very different domains, over a long time.
I'm an engineering leader working across AI, supply chains, and security. Twenty-five years of it, mostly in places where being wrong is expensive:
- Cisco. Founding engineer on a physical security and communications platform for defense and public safety, with surveillance cameras and push-to-talk endpoints managed as one fleet. Later CTO after it spun out.
- Illumina. Led the software behind DNA sequencers whose results feed clinical decisions.
- Jazzware. Owned the full AI mandate: strategy, product, and the hosting and inference infrastructure underneath it.
- Earlier. Research at MIT's AI Lab and DARPA-funded work, back when AI still meant rule-based systems. Five US patents along the way.
Now I run SurroundApps, a US company with an engineering team in Dhaka. We build verification infrastructure for industries where trust used to run on faith: garment supply chains, charitable giving, home healthcare, and device security.
I also publish open-source work on agent provenance, offline credential verification, and integrity monitoring. Running code, not position papers.
I grew up in Dhaka and I'm based in Silicon Valley.
New essay · October 2026
Nobody Told It Not ToIn 1982 an AI program won a fleet wargame by sinking its own damaged ships, because the rules never said it couldn't. This year, agents doing routine research went looking for ways into government data sites. On why guardrails can't just be rules, and whose job it is to write them.
Essays
Standalone pieces on AI and trust, newest first.
An AI program sank its own ships to win a wargame in 1982. Agents are now finding their own way into government data sites. On why guardrails can't just be rules, and whose job it is to write them.
A micro-factory making clothes on demand looks simpler and more local than the giant plant, but it doesn't make its own fabric, thread or dye. It assembles them from dozens of sources, mixed differently for every order. Software hit the same problem when apps started being built from other people's code, and learned that you can't trust a parts list where every supplier vouches for itself. On what physical supply chains can take from that before they repeat it.
Series
Two finished series that come at the same question from different altitudes: what makes an answer trustworthy, and what it takes to build one where the customer actually is.
The Verification Series
One argument in seven essays: how trust works when generation is free and verification is the scarce thing.
Forward Deployment
Closing the gap between generic software and one particular company is a thirty-year-old job. These essays are about the structure around it.
Elsewhere
I write the practitioner pieces, on cybersecurity and supply chains, at SurroundApps Perspectives. Recent op-eds: Your Phone Number Is Now Your Identity. Guard It in the Dhaka Tribune, and Digital Passports for Every Garment Could Reshape Bangladesh's Export Trade in bdnews24.com.

