Blockchain2024
NFTATTRIBUTIONENGINE
99.8% attribution accuracy across 1M+ daily NFT trades via raw 32-byte EVM word parsing.
99.8%
NFT attribution accuracy
1M+
Trades indexed per day
3
Marketplace protocols
SCROLL
99.8% attribution accuracy across 1M+ daily NFT trades via raw 32-byte EVM word parsing.
Standard ERC-721/ERC-1155 Transfer events reveal no marketplace or pricing data. Anonymous transfers needed to be attributed to marketplace orders across OpenSea Seaport, Blur, and X2Y2.
A Rust log correlation engine pulling full receipts, parsing raw 32-byte EVM words, matching Transfer events to marketplace signatures. Results normalised into unified schema with composite PostgreSQL indexes for sub-second queries on 10M+ records.
Build a correlation engine that parses full transaction receipts, iterates sibling logs using raw 32-byte EVM word parsing, and stitches anonymous transfers to platform-specific order events at 1M+ trades/day.
Pull complete transaction receipts for all NFT-related transactions, not just the Transfer event.
Iterate all logs in the same transaction, identifying marketplace-specific event signatures alongside the Transfer.
Parse marketplace order data from raw 32-byte EVM words — direct byte manipulation in Rust, no ABI decoder.
Normalise OpenSea, Blur, and X2Y2 event structures into unified production schema for cross-marketplace analytics.
Across OpenSea Seaport, Blur, and X2Y2 after correlating anonymous Transfer logs with marketplace order events — validated against spot-checked on-chain bundles so Messari-tier consumers could trust venue and price fields.
Across OpenSea Seaport, Blur, and X2Y2 after correlating anonymous Transfer logs with marketplace order events — validated against spot-checked on-chain bundles so Messari-tier consumers could trust venue and price fields.
Sustained real-time processing with a sub-2% error budget on noisy mempool-adjacent receipts, using Rust-side batching and composite indexes so dashboards stayed fresh without re-scanning history nightly.
Sub-second analytical queries on 10M+ normalized rows via composite PostgreSQL indexes tuned to marketplace + token + block clustering patterns common in NFT analytics workloads.
OpenSea Seaport, Blur, and X2Y2 share one production schema for trades, fees, and wash-risk signals — simplifying downstream GraphQL and BI models that previously forked per marketplace.
Marketplace contracts use non-standard encodings; ABI decoders can fail silently on edge cases while raw 32-byte word parsing keeps every branch explicit.
99.8% accuracy required encoding-variant tables per protocol, not a single generic decoder.
Zero-cost loops over logs and receipts without GC pauses during high-frequency ingestion windows.
Ownership and explicit error types made it safe to refactor handlers when Seaport shipped minor event tweaks.
Query paths for “latest sales by collection” and “wallet PnL” mapped cleanly to composite keys instead of ad hoc JSON blobs.
Normalized rows let BI tools join attribution output to pricing engines without another ETL hop.
Pulling full receipts (not just Transfer topics) unlocked sibling logs that pure topic filters would miss.
This was the difference between “some NFT moved” and “this blur bid settled at price X.”
We take on a small number of projects at a time. If the problem is hard, we're interested.