CRMS — Nepal Customs' Machine-Learning Risk Targeting Now Covers 90% of Commercial Trade
Nepal
Nepal Customs built an in-house machine-learning module inside ASYCUDA World, now running in the six largest customs offices, screening ~90% …
Mongolia · Ulaanbaatar · See the Mongolia profile · See the Ulaanbaatar profile
Top 66% 53/100 · Ask Evidence Copilot about this practice
Mongolia's Customs General Administration deployed HYPERGLORY, an AI chatbot using retrieval-augmented generation to recommend correct HS commodity codes to importers and exporters, built with a South Korean consortium under a government-to-government AI technology-transfer progr
Mongolia's Customs General Administration (MCGA) signed a joint cooperation agreement in Ulaanbaatar on 13 May 2025 with a South Korean consortium (Bituen, Virnect and Rebellions), funded through South Korea's NIPA AI-Semiconductor Overseas Expansion Support Project, aiming to reduce HS-code misclassification and the complaint backlog it caused.
The resulting system, HYPERGLORY, combines retrieval-augmented generation with a fine-tuned large language model running on Rebellions' NPU hardware to recommend HS commodity codes from natural-language product descriptions; it is billed as the first commercial NPU-based, rather than GPU-based, LLM/RAG deployment inside a live customs administration.
Year 1 of the two-year rollout was completed on 23 December 2025. The system obtained KTR AI certification and GS Grade 1 software-quality certification plus a public-institution priority-purchase designation, and the vendor reports a 2.7x power-efficiency advantage over GPU-based alternatives.
No independently audited accuracy or error-reduction figures have been published by MCGA itself; the strongest technical claims currently originate from the hardware vendor, Rebellions, rather than an independent auditor or the customs administration.
Read the full analysis: https://rebellions.ai/ai-powered-hs-code-recommendation-chatbot-for-mongolian-customs-built-on-llm-rag/
Implementation detail (cost, timeline, staffing, conditions for success) is not yet available for this practice.
Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.
Where this practice's information was retrieved from, and when.
Nepal
Nepal Customs built an in-house machine-learning module inside ASYCUDA World, now running in the six largest customs offices, screening ~90% …
Iraq
Iraq's Customs Authority deployed UNCTAD's ASYCUDA World, whose newest generation adds AI-based risk-scoring to sort shipments into inspection channels. Revenue …
Kazakhstan
Kazakhstan's Ministry of Finance built Smart Data Finance, an AI platform fusing tax, customs and treasury data into unified profiles …
Bulgaria
In March 2026 Bulgaria's NRA deployed BgGPT—a sovereign open-weight LLM—across central office and seven territorial directorates for 7,000+ staff; NRA …
Open full copilot Grounded in cited practices — always check the sources.