I spent years building enterprise AI at Google Cloud, then co-founded a company that was acquired for hundreds of millions. Now I'm leading AI strategy at Uniphore, building the platform that turns enterprise intelligence into a compounding advantage. I write about what I learn along the way.
Cupertino, CA · Building at the intersection of AI and enterprise
Early exposure to the hardware that would power the AI revolution. Learned to think at the intersection of silicon and software.
Built AI systems at scale. Learned how large organizations actually buy, deploy, and depend on AI. The pattern recognition that later became my unfair advantage started here.
Where I learned to think in decades instead of quarters.
Left Google to build. Took the enterprise AI thesis and turned it into a product, a team, a company. The thesis worked.
I've never been more excited about what I'm building. Open-weight models will capture 70–80% of the enterprise AI market — and Uniphore is positioned to own the stack that makes them production-ready. Stacked SLMs, autonomous agents, and a platform that lets enterprises build custom AI applications on their own data. This is the generational opportunity I've been building toward for the last decade.
Long essays for deeper theses, short posts for sharper observations.
AI is becoming more expensive to create and cheaper to use. The next advantage will come from turning the same machine intelligence into more verified value.
Why the next software war will be won at the improvement layer.
Top SLMs are already capable enough for a growing share of enterprise workflows. The next moat is the system that gives those models the right goal, memory, workflow state, constraints, and validation at each step.
A new paper reframes hallucination and overconfidence as symptoms of unbounded autonomy — and proposes a four-state framework (SMARt) for governing when agents can and cannot act.
WorkstreamBench tested agents on real financial modeling tasks. Claude's general web interface scored 69.1. Its specialized Excel add-in scored 60.4. Same model, different harness, 3.8× gap.
SKILL0 trains small models with curated skills then progressively withdraws them. The result: a 3B model scores 87.9% where GPT-4o scores 48.0%, at 5× lower inference cost.
The complete rethinking of how enterprises build and deploy AI agents. Stacked SLMs, a self-improving flywheel, and an architecture where every interaction makes the system smarter.
Orchestrated small language models that beat frontier models at a fraction of the cost. The routing layer is the real product.
Real-time conversation intelligence across every customer interaction. Emotion detection, agent guidance, and insights that turn conversations into competitive advantage.
AI agents that activate customer data to drive personalized engagement, orchestrate campaigns, and maximize ROI across every channel. Built on an AI-first customer data platform.
One platform connecting knowledge, SLMs, agents, and trust infrastructure. Where the SLM Flywheel, agent architecture, and enterprise governance come together as a single system.
The enterprise software market is about to be rebuilt. Platforms that let companies create custom AI applications — not just use pre-built ones — will dominate. Uniphore is where I'm putting everything I've learned into building that platform.
I believe open-weight models will own 70–80% of the enterprise market. The winners won't be the companies with the best foundation model — they'll be the platforms that make those models deployable, governable, and compounding. A proven exit, deep enterprise AI experience, a Stanford network, and trilingual access to the American, Chinese, and Korean markets — all of it is deployed at Uniphore. I write about enterprise AI because I'm building it every day.
I publish on LinkedIn and update this site as I build.