A novel framework that integrates large language models with traditional operations research optimization — enabling natural-language interaction with complex supply chain decision systems.
Filed July 24, 2025 — Powerlab Inc., Cambridge, Massachusetts
Traditional supply chain optimization tools are powerful but inaccessible. They require deep expertise in mathematical programming, simulation, and data science to configure, run, and interpret. This creates a gap: the people who need the insights (supply chain managers, importers, compliance teams) often can't use the tools that produce them.
The patented system bridges this gap by creating an "agentic" framework where large language models serve as an intelligent interface to complex optimization engines. Users interact in natural language — asking questions, exploring trade-offs, and receiving explanations — while the system translates those interactions into formal optimization problems, solves them, and presents results in plain language.
"What happens to my landed cost if I switch this product from China to Vietnam?"
The system understands the question, identifies the relevant product, looks up the correct HTS codes, calculates all applicable tariff layers for both countries (MFN + Section 301 + Section 232 + Section 122 + MPF + HMF), compares the totals, and responds with a clear explanation of the cost difference and trade-offs — in seconds.
LLM-powered conversational layer that translates business questions into formal optimization problems and explains results in plain language.
Traditional operations research methods — linear programming, mixed-integer programming, network flow optimization, and simulation — solving real supply chain problems.
Intelligent middleware that decomposes complex queries into sub-problems, coordinates between the LLM and optimization modules, and maintains conversational context.
Enables "what-if" exploration: users modify constraints, parameters, and assumptions conversationally and see updated optimization results instantly.
Every recommendation comes with reasoning: why this option, what are the trade-offs, what constraints are binding, and what would change the answer.
Optimizes supply chain network configuration: facility locations, sourcing decisions, transportation routes, and inventory positioning across multi-echelon networks.
The Trade Lab is the first commercial application of this patented technology, applied specifically to trade compliance and customs operations. The same agentic framework that enables conversational supply chain optimization powers:
The system analyzes product descriptions, photos, and specifications using the agentic pipeline — coordinating between vision models, web research, tariff databases, and CBP rulings to produce HTS classification suggestions.
The optimization engine identifies the lowest-cost compliant import strategy — considering tariff layers, FTA eligibility, country-of-origin rules, and exclusion programs to minimize total landed cost.
The natural language interface enables users to ask compliance questions conversationally — "Does my product qualify for USMCA?" or "What's the total duty on steel pipe from India?" — and receive accurate, sourced answers.
Title
AI-Powered Supply Chain Intelligence System
Application Number
63/849,878
Filing Date
July 24, 2025
Type
U.S. Provisional Patent Application
Inventor
Kevin Power
Assignee
Powerlab Inc.
Specification
35 pages + 15 pages of drawings
Status
Patent Pending