SpecOS: AI Product Data Intelligence Platform

SpecOS is a product-data intelligence platform for distributors. Customers search 328K+ construction products in natural language or by photo, and every unanswered query becomes a structured data gap the system can resolve.

I designed and built SpecOS end to end: the data engine, search product, administration tools, identity, and public site. The platform turns a large distributor catalog into a searchable, cited product knowledge system.

Customers ask questions in natural language or upload a jobsite photo. SpecOS searches 328K+ products, extracts specifications from documentation, and returns answers with citations. If the evidence is missing, it says so instead of inventing a specification.

Every unanswered question becomes a data gap tied to a product category and requested feature. SpecOS searches the distributor’s documentation, proposes an enrichment with source attribution and confidence, and either applies it automatically or routes it for review.

Search combines Pinecone with a Postgres catalog containing 328K products, 1.25M variants, and 9.2M features. Thirteen specialized tools sit behind one interface, while a model-agnostic layer supports OpenAI, Gemini, or Claude.

SpecOS turns customer questions into a measurable feedback loop for catalog quality. High-confidence gaps can be resolved in under 30 seconds, every answer is cited, and every AI interaction is logged for compliance.