How JD Group Reduced Tariff Classification from Hours to Milliseconds with AI

About JD Group

JD Group is a leading binational logistics company specializing in comprehensive customs and logistics services for the manufacturing and trade industry. With over 25 years of experience and more than 500,000 operations per year at 99% accuracy, they serve clients across the US-Mexico border corridor.

Their presence spans Tijuana, Ensenada, Mexicali, Tecate, Manzanillo, San Diego, and Calexico, with strategic alliances in Nuevo Laredo, Ciudad Juárez, Nogales, Mexico City Airport, and Guadalajara.

The Challenge

JD Group's biggest operational bottleneck was tariff classification — the process of assigning the correct customs code to every product that crosses the border. With a high daily volume of classifications handled by a team of 20 human classifiers, classifying a single new product could delay the entire operation line anywhere from 30 minutes to 4 hours.

The stakes were enormous. A single incorrect tariff code doesn't just delay a client's operation — it can result in fines of hundreds of thousands of pesos from Mexican and US authorities. To keep scaling and maintain their industry leadership, JD Group needed to eliminate this bottleneck by adopting Machine Learning and AI.

Our Approach

YUNO built a custom Machine Learning classification model leveraging JD Group's most powerful competitive advantage: 25 years of pre-classified tariff data. This proprietary dataset — something no competitor had access to — enabled us to train a model that delivers prioritized classification options ranked by confidence percentage in milliseconds.

• — Trained a supervised ML classification model on 25 years of historical tariff data
• — Built a REST API so any system in the organization can classify products instantly
• — Created a custom chatbot interface for conversational classification queries
• — Designed embeddable screens for integration into JD Group's existing enterprise systems
• — Deployed entirely on JD Group's private infrastructure — zero data shared with third-party LLM providers
• — Built continuous learning pipeline so the model improves with every new classification

Data Sovereignty

A critical requirement for JD Group was absolute data privacy. Their 25 years of classified tariff data represents their most important competitive advantage going forward. The entire solution runs on JD Group's private infrastructure, and none of their data is shared with any LLM provider like OpenAI or others. This ensures their proprietary knowledge remains exclusively theirs.

The Results

While competitors had to raise $5 million in financing to develop a similar prediction model — one that can take minutes per classification — JD Group leveraged their 25 years of proprietary data to build a more accurate model at a fraction of the cost. The same team of 20 classifiers can now process hundreds more products per day, dramatically increasing the profitability of every operation. Anyone in the organization can now classify goods through the API, chatbot, or directly within their enterprise systems.

"Protecting our data was non-negotiable. With YUNO we achieved classification certainty in nanoseconds — that's critical for avoiding costly fines and keeping our clients' operations moving without delays."

— Jonathan Canales, Director of IT, JD Group