Loomi
Natural-fiber clothing marketplace with product-level material transparency.
Problem
Natural-fiber products are scattered across dozens of stores with inconsistent, buried material data; no tool gives standardized, filterable product-level fiber composition.
What I built
A marketplace that ingests catalogs from 10+ retailers, parses and classifies fiber composition, cross-references affiliate networks, and serves a unified shopping experience with 'Fabric Facts' transparency panels. A two-tier data architecture separates raw JSONB staging from a normalized serving layer so the pipeline enriches data without blocking reads.
Engineering highlights
Concurrent Go scraping (goroutines, cursor pagination, SSRF protection, rate limiting); a fiber parser handling 6+ composition formats; quality gates that promote records only after affiliate verification; a 9-module Go architecture; REST API with Redis caching; a self-hosted WebP image pipeline; PostgreSQL with GIN indexes and 14 migrations. 765+ products live across 10+ brands.