datum
Bringing TradFi-style closing auctions to crypto with Canton privacy.
Project Description
Datum is building decentralized closing auctions for crypto markets on Canton. In traditional finance, closing auctions are critical liquidity events used by ETFs, pension funds, asset managers, and traders to execute large orders at a single market-clearing price. These auctions determine the official closing price used for valuation, benchmarking, and portfolio rebalancing.
Crypto lacks an equivalent mechanism, forcing institutions to trade in fragmented markets with significant market impact and information leakage. Datum introduces scheduled daily crypto auctions aligned with APAC, European, and US market closes. Participants submit orders during a fixed auction window, which are then matched at the price that maximizes traded volume.
Canton's privacy and selective disclosure capabilities make this possible. Orders remain confidential during the auction, preventing front-running and information leakage (which is a blocker for auctions in crypto due to being continuous markets), while selective views allow the same settlement to be displayed differently to different audiences. The result is a trusted closing price, improved liquidity, better execution quality, and institutional-grade market infrastructure for digital assets.
How it's Made
We made four Daml contracts on Canton DevNet. All three privacy tiers come from a single mechanism so "public," "regulated," and "confidential" are the same code path with different audiences, not three separate systems. Authorization is fully declarative: a TradingAuthority contract carries the owner's signature, so when the exchange acts on it the transaction inherits the owner's authority without any key changing hands.
Off-ledger, a mvp using a Python/Flask call-auction engine uncrosses orders at a single clearing price and hands the result across a fixed MatchResult seam to a zero-dependency (stdlib urllib) client for the Canton JSON Ledger API v2. An explorer makes the privacy model visible.
The UI is deployed via Vercel and links to the matching engine api which is deployed via AWS.
