Field CISO · AI Builder

Usman
Chaudhary

The advisor who ships. Advising Fortune 500 enterprises and federal agencies on AI-era security — and building the working systems that prove what's possible.

I'm a Field CISO and AI Security Executive at Google Cloud, advising CISOs and boards on AI governance, SecOps transformation, and sovereign AI strategy. Previously at Palo Alto Networks as Business Value Director, and Accenture and KPMG as a strategy consultant.

My background spans deep engineering, management consulting, and enterprise security GTM: MS Electrical Engineering, University of Michigan (research in secure vehicular communications) · MBA, University of Chicago Booth (Dean's Honor List) · 9 years at Ericsson architecting hybrid cloud and IoT security infrastructure.

In 2026 I moved from advising on AI to building with it: working systems for the problems my customers actually face — agentic workflow automation under enterprise constraints, retrieval for regulated environments, end-to-end fine-tuning infrastructure, and AI-native security for disconnected networks. I speak at Billington CyberSecurity, GovExec, and ATARC, and write on what it takes to build genuinely AI-native security programs.

Usman Chaudhary
ROLEField CISO, Google Cloud
LOCATIONWashington DC Area
EDUCATIONBooth MBA · Michigan MS EE
EXPERIENCE

Career

Google Cloud
2022 – Present
Field CISO & Strategic Growth Executive
Trusted advisor to CISOs and C-suite across Fortune 500 and federal agencies on AI-era threat posture, SecOps modernization, and sovereign AI strategy. Speaker at Billington CyberSecurity and GovExec. Developed Legacy SOC Displacement framework and AI security training for 300+ sales reps globally.
Palo Alto Networks
2021 – 2022
Business Value Director
Engineered commercial frameworks resulting in the two largest contracts in company history. Advised Fortune 500 boards on cybersecurity and digital transformation investment decisions.
Accenture
2019 – 2021
Business Strategy Manager
Multi-year digital transformation strategies for Fortune 500 across media, telecom, and technology. Built joint B2B GTM strategy for a SaaS IoT product identifying enterprise AI and edge computing market trends.
KPMG
2017 – 2019
Senior Strategy Consultant
Performance improvement and GTM strategy for healthcare and industrial clients. Dealer profitability model for a Global Automotive OEM resulting in multi-million dollar implementation.
Ericsson
2008 – 2017
Technical Product Manager / Product Architect
9 years leading hybrid cloud and NFV architecture for major Tier-1 US telecom. Established client's IoT Innovation Lab and produced 6-year infrastructure security forecast.
AI Systems

Working systems, real constraints

Every system here starts with a problem enterprises actually have — and a constraint they actually face: enterprise OAuth boundaries, regulated-data retrieval, sovereign deployment, zero network egress. Designed, built, and shipped end-to-end.


P6 — EdgePatch v1 — complete · P7 — Open source + publication — next
P1MAY
2026
AI Chief of Staff
Executive workflow automation under enterprise OAuth constraints
Reads a live Gmail inbox and Google Calendar, prioritizes the day, and triages email with specific recommended actions — powered by the Claude API behind server-side OAuth 2.0, the same trust boundary an enterprise deployment has to respect.
Claude APIGmail APIGoogle CalendarOAuth 2.0PythonStreamlit
P2MAY
2026
AI Content Engine
Editorial intelligence pipeline with human-in-the-loop publishing
Synthesizes a curated reading list into structured intelligence records, then drafts Tier 1/2/3 content — full articles, LinkedIn posts, quick reactions — in a consistent voice. Firestore-backed library with a persistent action queue, so a human approves everything before it ships.
Claude APIFirestoreGoogle SheetsPythonStreamlit
P3JUN
2026
Frontier AI Intelligence RAG
AI intelligence assistant for a regulated SOC — answers grounded in source citations
Hybrid BM25 + vector retrieval over 30+ primary documents from frontier labs and government bodies — Anthropic, OpenAI, Google DeepMind, Mistral, DeepSeek, Meta, CISA, UK AISI. Every answer cites its sources, the requirement that separates a demo from something a compliance-bound team can actually use. Field CISO mode, live model leaderboard, auto-watcher.
Claude APICloud SQLpgvectorVertex AIPythonStreamlit
P4JUN
2026
UrduGPT — LLM from Scratch
Depth proof — transformer internals, built and inspected by hand
Character-level transformer built from scratch on 290M Urdu characters: 11M parameters, trained on Apple Silicon in 41 minutes. Attention visualization shows the layers specializing — Layer 0 tracks character sequences, deeper layers form word-level patterns — an interpretability case study on a low-resource language, with an honest qualitative comparison against Gemma 12B.
PyTorchTransformersUrdu NLPInterpretabilityPython
P5JUN
2026
Urdu Fine-tuning Pipeline
End-to-end fine-tuning infrastructure: local prototype → cloud production pipeline
The same customer fine-tuning workflow on two stacks: LoRA on Apple Silicon with MLX for fast local iteration, then a containerized GCP training pipeline — Docker, Artifact Registry, GPU training on Vertex AI — taken end-to-end on real cloud infrastructure. At a matched training budget, the cloud run's deeper 16-layer adapter on Mistral 7B reached a notably lower evaluation loss than the leaner local prototype — an honest before/after on what fine-tuning a small model actually buys.
MLX-LMHuggingFace + PEFTLoRAVertex AIDockerEvals
P6JUN
2026
EdgePatch — Offline Patch Checking
Checking AI-written C security patches offline, before they ship
A tool for checking AI-written security patches without an internet connection. An AI suggests a fix; EdgePatch runs a fixed set of automated checks — does the patch apply, does it change the right place, does it build and pass the tests — and packages the results into one report for a person to approve. The AI writes the patch; the checking doesn't rely on the AI. v1 includes a benchmark across four real C libraries (zlib, libpng, expat, libxml2) that anyone can reproduce with a single command, fully offline. Built for critical infrastructure, industrial control, and other high-stakes systems the cloud can't reach.
PythonOffline / air-gappedAutomated checksReproducible benchmarkLocal LLM
P7SEP
2026
Open Source + Publication
Putting the work where others can run it, test it, and break it
Open-sourcing EdgePatch with full benchmark results against real CVEs, publishing the findings, and contributing upstream to the tooling these projects are built on.
OSSBenchmarksResearch writeup
ROADMAP

2026 Build Roadmap

May 2026
Agent Architecture
APIs, OAuth, Gmail, Calendar
✓ DONE
May 2026
Content Engine
Claude API, Firestore
✓ DONE
Jun 2026
Frontier AI RAG
pgvector · Vertex AI · hybrid search
✓ DONE
Jun 2026
LLM from Scratch
11M-param transformer · interpretability
✓ DONE
Jun 2026
Fine-tuning Pipeline
MLX LoRA local · GCP GPU cloud
✓ DONE
Now
EdgePatch
Offline patch evaluation · v1 shipped
● NOW
Sep+
Open Source
OSS release · publication
Writing

AI, Security & Ideas

Original analysis on AI-era threats, autonomous attack surfaces, and what it means to build genuinely AI-native security programs.

RESEARCH

Academic Publications

Speaking

Keynotes & Panels

Federal cybersecurity forums, AI security summits, and government technology conferences. Speaking at the intersection of AI, Zero Trust, and enterprise security modernization.

MAY 2026
Government Innovation Summit — Pennsylvania
Public Sector Network · Keynote
"AI & Cloud Innovation for Mission Impact: Secure, Scalable Solutions for the Public Sector"
APR 2026
Cloud Security Summit — Reston, VA
GovExec · Fireside Chat
"Securing your Agency's AI and Data Modernization Journey" — how agencies can deploy AI securely using Google's Secure AI Framework (SAIF)
FEB 2026
Cyber Futures Summit — Reston, VA
ATARC / Nextgov / FCW · Fireside Chat
"Building an AI-Powered Zero Trust Defense: Modernizing Government Security Operations and Empowering the Future Workforce"
AUG 2025
GovExec Webinar — Modernizing Security Under Pressure
GovExec · Webinar
How federal security teams are modernizing SecOps under resource, talent, and compliance pressure using AI-augmented detection and response.
2025
Billington CyberSecurity Summit
Billington · Panel
"MXDR Strategies for Federal Defense" — AI-powered managed detection and response for federal cybersecurity programs.
Perspectives

Policy, Economics & Social Good

Long-form analysis on technology's impact on emerging economies, energy transitions, and social systems. Data-driven, independently written, intentionally contrarian where the evidence demands it.

Energy · Economics · Pakistan
Running on Sun: How Pakistan's Consumer Revolution Could Break the Oil Trap Forever
Pakistan has accidentally launched one of the fastest energy transitions in the developing world — entirely without government planning. 27–33 GW of consumer-deployed solar, $12B already saved in oil imports, and a credible path to energy sovereignty by 2040. A data-driven analysis with full source citations.
APR 2026 Read ↗
Connect

Let's connect

Security executive who builds — exploring what happens when deep domain expertise meets hands-on engineering.

Particularly interested in conversations around enterprise AI security, agentic AI risk, retrieval and fine-tuning for regulated environments, and bringing AI-native security to disconnected and sovereign networks.

If you're building in these spaces, advising enterprises on AI adoption, or just want to compare notes on what you're seeing in the field — reach out.