Hacker

kamal singh

@kamalbuilds

Joined Sep 2026Software Enginner at DELHI TECHNOLOGICAL UNIVERSITYIndia

Projects

  • Top 50

    Undertow

    Graph HacksOct 2026

    When a digital dollar breaks, the losses hide behind layers of repackaging. Undertow finds where they land. DeFi savings vaults lend depositors' money to borrowers who lock up a guarantee, usually a digital dollar like USDe. That guarantee is rarely the plain token: it is wrapped into a savings version (sUSDe), then a fixed-term version (PT-sUSDe), and vaults even lend into other vaults. When the underlying dollar drops, warning systems that search by name miss every repackaged layer. Undertow loads live Morpho mainnet into FalkorDB (252 vaults, 220 markets, 156 assets, 29 curators, 561 allocation edges, 53 on-chain-verified wrapper edges) and follows a price shock hop by hop: through wrappers, into lending markets, up vault-of-vault allocations, and through vault share tokens reused as guarantees elsewhere. On today's data, a 30% drop in USDe: a name search flags 2 vaults and $10.6M; Undertow finds 10 vaults and $31.1M. For USDS, the name search finds nothing; Undertow finds 19 vaults and $20.5M. A tool-calling agent works the same graph: it runs shocks, explains the exact path behind every number, and drafts the curator's response (cap the exposed market, reallocate into the largest same-loan market whose guarantee sits outside the shocked asset's wrapper family). Every Cypher statement it runs is shown in the UI. Every shock is stored as a Scenario in the graph, and each chat session lives in its own FalkorDB graph.