Overview
An AI system that sorts websites into IAB taxonomy categories for adtech and marketing, served as real-time APIs.
The problem
Rule-based classification falls apart at scale. Content is ambiguous, categories overlap, and coverage across the whole web is uneven.
What I built
Fine-tuned language models with prompt-tuned classification logic, sitting behind stateless, containerised FastAPI endpoints.
- Clean categories. Output maps to IAB categories, not loose free text.
- Fast serving. Stateless services and Docker deployments that scale the cloud-native way.
- Disambiguation. Extra logic to handle overlapping and ambiguous content.
Result
Accurate, scalable categorisation that lines up with the industry standard.