Loading...
Loading...
2024 · AI project
Guillaume Schonrock · Updated on
AI-assisted customs classification based on hybrid retrieval and explainable outputs.
Tariff8AI combines hybrid retrieval and AI to produce explainable customs classifications. Indexing and semantic search are components of this same project.
I scoped the classification logic and built the RAG pipeline, FastAPI backend and workflow for consulting recommendations.
Indexed customs knowledge and explainable classification
≈ 50,000 indexed entries
Hybrid
Explainable
Customs classification relies on a code corpus and business rules. The goal is to retrieve relevant information and make recommendations understandable.
The pipeline combines FastAPI, LlamaIndex, Qdrant and the OpenAI API. Retrieval uses keywords and semantic similarity before producing a classification with explanations.
The corpus contains approximately 50,000 entries across several customs-code indexes. A code may appear in multiple indexes: this total does not count documents or unique codes.
No measured accuracy rate, response time or time savings are claimed.
Let's talk about your product, platform, or operational workflow and see how I can help shape and build it.