Trademark Guard

The hard part

Merchant catalogues arrive in bursts of tens of thousands of items, and every tenant's data has to stay provably invisible to every other tenant. The answer was an event driven pipeline behind a Redpanda bus with Celery workers, and isolation enforced in PostgreSQL itself through Row Level Security rather than in application code that can be forgotten.

The problem

Sellers on Amazon and Walmart carry catalogues far too large to police by hand, and a single trademark conflict can pull a listing down. Checking every product title and brand against the USPTO register is a continuous job, not a one time audit.

The approach

I built an event driven microservices system where catalogue import, normalization, and conflict matching are separate stages connected by a message bus, so a slow import never blocks matching and a failed match never loses the record.

What it does

• Amazon SP-API and Walmart store connectors with catalogue import
• Continuous trademark conflict monitoring against USPTO registrations
• Bidirectional fuzzy matching engine with confidence scoring
• Strict per tenant data isolation through PostgreSQL Row Level Security
• Single command deployment of the full stack via Docker Compose

Architecture

A Next.js 15 frontend over a FastAPI service layer. Celery workers consume from a Redpanda message bus for import and matching stages, Redis backs the queue, and PostgreSQL holds tenant data behind Row Level Security. Elasticsearch serves the matching engine, with a designed upgrade path to native kNN vector search.

Outcome

Merchants get continuous conflict detection instead of periodic manual audits, and the tenant isolation model is enforced at the database layer rather than trusted to application code.