Blockchain2023
DEFISUBGRAPHS
Production subgraphs for 6 DeFi protocols. 100+ institutional clients. Sub-2% error rate.
6
DeFi protocols shipped
100+
Institutional downstream clients
<2%
Reconciliation error budget
SCROLL
Production subgraphs for 6 DeFi protocols. 100+ institutional clients. Sub-2% error rate.
Production subgraphs for GMX, Livepeer, Mummy, Ellipsis, Level, and Ribbon protocols serving institutional clients.
TypeScript/AssemblyScript subgraphs with on-chain validation, reconciliation, and open-source SDK reducing dev lifecycle 40%.
Normalise heterogeneous protocol events into standardised schema for cross-protocol analytics at sub-2% error rate.
Pick Messari schema families (AMM, yield, perps, generic) per protocol and lock entity boundaries early.
AssemblyScript handlers decode ABIs, normalize decimals, and emit financial snapshots.
Mustache-templated configs for addresses, start blocks, and RPC endpoints per chain.
Compare subgraph totals to on-chain explorers and internal spreadsheets before client-facing release.
Upstream reusable helpers for shared patterns (e.g., forked perp protocols) to speed sibling subgraphs.
GMX, Livepeer, Mummy, Ellipsis, Level, and Ribbon each shipped with Messari-standard entity shapes so downstream analysts could compare perp volume, staking flows, and AMM liquidity with one mental model.
GMX, Livepeer, Mummy, Ellipsis, Level, and Ribbon each shipped with Messari-standard entity shapes so downstream analysts could compare perp volume, staking flows, and AMM liquidity with one mental model.
Messari’s paying desks consumed these subgraphs through the same GraphQL contracts as the rest of the catalog — uptime and schema drift were treated as product issues, not one-off integrations.
On-chain reconciliation scripts and manual spot checks held financial metrics within institutional error budgets before manifests promoted to production networks.
Open-source SDK and template work shrank the average subgraph iteration loop — less copy-paste across forks meant more time on protocol-specific edge cases.
WASM handlers stay close to chain semantics while TypeScript tooling supports codegen and tests.
Schema-first work prevented six protocols from diverging into six custom JSON blobs.
Parameterized deployments scaled to many networks without maintaining duplicate subgraph.yaml trees.
Error budgets stayed sub-2% because promotion gates included on-chain reconciliation, not just CI green.
We take on a small number of projects at a time. If the problem is hard, we're interested.