Track record
Delivered where reliability isn’t optional.
Fifteen years of building and selling AI inside organisations where a wrong answer has consequences. Named references are available under NDA.
Lead engagement
US defense institution
Mission-critical AI systems, delivered under constraints most commercial deployments never encounter.
The requirement was not accuracy in the abstract, it was reliability, security and precision under conditions where a wrong output is not a bad recommendation but an operational failure. That forced decisions in the architecture: deterministic guardrails around probabilistic components, full auditability of every inference, and a deployment model that assumed constrained connectivity rather than a cloud round-trip.
The same discipline is what Thimar applies to statutory filings today: a language model proposes, a rules engine verifies, and nothing reaches a regulator without a check that can be shown.
Outcome
| Engagement | Withheld |
| Scale | Withheld |
| Outcome | Withheld |
Client naming and outcome metrics pending permission . PLACEHOLDERS D4, D7. Named references available under NDA.
The record
Five threads, run in parallel.
These are not sequential roles. They overlapped, which is the point, the commercial work funded the technical work and vice versa.
AI & ML leadership
Enterprise BD
CTO / architecture
Serial entrepreneur
Author & educator
15+ YEARS
From early NLP and computer vision through to today’s multimodal, agentic systems, architected production AI at global scale.
FORTUNE 50 & DEFENSE
Multi-million quota ownership and eight-figure pipeline closed with industrial and defense institutions.
MULTIPLE VENTURES
Hands-on architecture across NLP, deep learning and LLMs, using Python, PyTorch, AWS, GCP and Azure.
3 COMPANIES FOUNDED
Zero to operating company, three times over, across product, GTM, fundraising and delivery.
2 BOOKS
Two published books on generative AI and DeFi for business leaders; educator to thousands of professionals.