Enterprise Knowledge Search
Healthcare workers in low- and middle-income countries needed current, evidence-based medical guidance despite limited connectivity, scarce localized content, language barriers, and heavy frontline workloads.
Reported impact
CMES-AI now supports more than 17,000 healthcare practitioners across 19 countries and improves care for 20 million people each year, while cutting query time from 5 minutes to 2 seconds and increasing token throughput 22x so clinicians can make faster decisions at the point of care.
Techies Without Borders built CMES-AI as a medical knowledge assistant that combines generative AI with NVIDIA edge and inference tooling to deliver answers inside resource-constrained clinics. The team began by training and fine-tuning clinical LLMs on an NVIDIA T4 Tensor Core GPU, then deployed the runtime to the NVIDIA Jetson Nano Developer Kit so clinics could run the assistant locally even when connectivity is poor. TWB added a preferred orchestration of NVIDIA NIM inference microservices and powered retrieval and response quality with NVIDIA Nemotron Rerank, Embed, and Instruct NIM microservices. That stack lets clinicians ask natural-language questions against a continuously updated knowledge base, receive evidence-based answers in their own language, and use the system as a practical point-of-care assistant instead of relying on slow manual search or tag-based workflows.
A clinician submits a natural-language question to CMES-AI, the platform retrieves relevant medical knowledge, NVIDIA Nemotron Rerank, Embed, and Instruct NIM microservices refine and answer the request through NVIDIA NIM, and the Jetson-based clinic deployment returns an evidence-based response locally.
By minimizing latency while maximizing referential accuracy, CMES-AI moved from “tag-based processing” to “natural language queries,” enabling caregivers to access richer insights at the point of care.
CMES supports over 17,000 healthcare practitioners across 19 low- and middle-income countries, improving care for 20 million people each year.
Powered by NVIDIA NIM™, query response time dropped from 5 minutes to just 2 seconds.
Powered by NVIDIA Nemotron™ Rerank, Embed, and Instruct NIM microservices, CMES slashed query processing time from 5 minutes to just 2 seconds.
Token processing speed also jumped 22x—from 0.83 to 18.23 tokens per second—delivering faster, richer medical insights.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
numberOfEmployees":{"value":10,"@type":"QuantitativeValue"}
Related
NVIDIA Sample Sep 30, 2026 EPRI EPRI needed to run advanced AI for sensitive utility documents and weather-planning workflows locally, because energy-sector research data could not leave controlled environments and shared compute delays limited how… 150 tokens/sec aggregate across 10 concurrent Power Chat users DGX Spark · Dell Pro Max with GB10 · NVIDIA CBottle 
IBM Sample Aug 14, 2026 Wikimedia Deutschland Wikimedia Deutschland needed to make Wikidata’s massive, multilingual knowledge graph usable for modern generative AI applications because the existing access pattern depended on specialized query languages and complex… 180 employees IBM DataStax Astra DB · IBM watsonx.data NVIDIA Sample Aug 3, 2026 StudyFetch StudyFetch needed to keep AI tutoring affordable while scaling live lecture transcription and personalized learning for millions of students, but off-the-shelf cloud transcription created cost and latency ceilings that… 100M+ learning interactions NVIDIA H100 GPUs · NVIDIA L40S GPUs · NVIDIA NIM 