AI and fintech product development

Holdex is a Hong Kong design and technology studio that helps founders ship production AI and fintech products, including regulated digital-asset products.

Founders come to us when the product has to reach production and the senior team to build it does not exist yet: an AI prototype that works in a demo, or a fintech product with a launch window that will not wait. We embed with the founding team, and the people who scope the work are the people who build it.

What do we build?

AI products

When the AI is the product, or a feature inside a product users already rely on, we build it end to end: architecture, retrieval, evals, and the interface, as one team with no handoffs. We also find the workflows where AI changes what operations cost, and we audit AI that shipped and is not performing. The AI page covers the four offers in full.

Fintech products

Financial products have to be right from the first transaction. We build digital-asset protocols, institutional fintech, the blockchain infrastructure both run on, and the interfaces users trust with real money. The fintech page covers the four offers in full.

What have we shipped?

Clearpool is an institutional credit protocol. We co-built it with its founders from 2021, and a Holdex squad still runs its smart contracts, dApp, and product engineering. Six weeks after mainnet, Jane Street and BlockTower opened a permissioned pool, and the institutional tier that followed launched with KYB checks and an audit on every release. The protocol has originated more than $900M in loans.

Ithaca is an on-chain options protocol. We came in at the founding. It launched on Arbitrum and closed Hashlock and Hacken audits clean, and we built the Telegram mini-game that feeds its user funnel. Ithaca has passed 50,000 active users.

Truflation runs a network for verifiable economic data. We build its application layer: the explorer, the SDKs, the smart contracts, and the prediction-market interface. We also contributed to the design and engineering of its open-source MCP server, which gives AI agents direct query access to that data. The SDK we lead powers integrations with Robinhood, Bitstamp, and Raydium.

How does an engagement work?

The senior people you meet before we start are the people who build the product. A typical engagement is two to four senior engineers and designers plus a co-founder or partner, with no project-manager layer between you and them. We start with strategy before code: we challenge the assumptions behind the roadmap, agree a plan with you, and ship it in the open. We work like co-owners of what we build, which is why a Holdex squad has run Clearpool's product and engineering since 2021.

Why work with us instead of building in-house?

Building in-house costs founder time: months of hiring before anything ships, then managing the team instead of doing the work only you can do. We bring a senior team that already works together, so the context lives with the people building from week one.

When are we the right fit?

Most founders we work with are between pre-seed and Series A, with a product that has to ship in the next six to twelve months. The engagements that go best have a founder who can make the calls in the room and users who need the AI or financial product being built.