Diana Seo — Oriya, and the layer beneath AI

Writing — a working ledger

A ledger of non-consensus bets — from early crypto and cross-border fintech to building in AI today.

The Control Plane a four-part series

Part 1 The Gateway Is Plumbing. The Control Plane Is the Prize.Why this layer exists: every layer of the AI stack, and why the durable money sits with the one that can say no. live Part 2 The AI Maturity LadderCompanies grow into needing a control plane, they don't start there. The ladder, and a ten-agent simulation of where a policy actually holds. live

Part 3 Everyone Agrees on Control. Nobody Agrees on Where It Lives.Mapping the four competing architectural bets for AI control: gateway, lifecycle, neutral layer, enterprise platform. A fair map each camp would recognize. in progress

Part 4 Why Enterprise AI Governance May Become Global AI Governance.Where the enterprise control layer and the coming public one converge. in progress

Work & track record

Currently building

◇ Oriya

I grew up between three countries and as many healthcare systems, none of which agreed with each other. Then I did the thing everyone recommends and bought the wearables — and ended up with five dashboards, four scores, and no read on what any of them actually wanted from me. A bad number could quietly cost me a morning. That's the problem I build against.

One fair 0–100 recovery score to compete on — how well you recovered, not whose device runs generous.

Oriya normalizes Oura, Garmin, Whoop, and Apple onto that single index. Private beta in SF & NYC; the long arc is health data you own.

Visit Oriya ↗

Diana has been building — and selling — things since she was fifteen: e-commerce, simultaneous interpreting for government events and investor demos, albums of her own classical compositions. She grew up between Guangzhou, Cairo, and Seoul — inside two of the fastest-changing economies on earth — watching who got to participate in that growth and who was quietly routed around it. That question has shaped everything she's built since.

She studied across USC, Bocconi, and HKUST, and passed on the offers that came with them — from JPMorgan, Bloomberg, and Bain — to build instead. She's worked both sides of the table since: an early hand on growth and product across a run of hard fintech and web3 companies — VegaX Holdings, Solana Pay, Bifrost and BiFi, Demether, Tashi — and inside venture at Nest VC across Hong Kong and Singapore.

She's scaled startups from seed to Series B, and backs founders building where most funds don't look — Accra, Lagos, Vilnius, Bali.

AI is making the product itself easy to copy. What it can't copy is distribution, trust, and a growth loop that compounds — the part I've spent my career on, in the hard version of fintech and web3 and now in AI. These days I take on a couple of fractional growth engagements at a time, and back the startups I believe in — sometimes with a check, always with the work: the loop that hasn't clicked, the developer tool individuals love but enterprises won't yet buy, the market your team can't reach, the inference bill outrunning the product. My edge is the growth most advisors can't do — cross-border, into the markets where the next users actually are. If you're teaching yourself into tech, that conversation's free — those I answer first.

my inbox is open ↗