NVIDIA and DataStax Revolutionize Multilingual AI with NeMo Retriever Microservices

NVIDIA Introduces NeMo Retriever for Multilingual AI

NVIDIA has unveiled its latest innovation in the field of artificial intelligence: NeMo Retriever microservices. This groundbreaking technology is designed to enhance multilingual generative AI capabilities, addressing the growing need for language-inclusive AI solutions in global enterprises 1

1

.

Key Features and Capabilities

NeMo Retriever offers several critical functionalities:

1. Multilingual Information Retrieval: The system can search, process, and retrieve knowledge across multiple languages, enabling AI to deliver more accurate and globally relevant outputs 1

   1

   .

2. Cross-Format Understanding: NeMo Retriever can comprehend information from various formats, including documents, making it versatile for diverse enterprise needs 1

   1

   .

3. Vector Embedding: The technology transforms data into AI-ready formats, significantly speeding up the processing of large datasets 2

   2

   .

4. Evaluation and Reranking: NeMo Retriever examines and ranks retrieved information based on relevance to queries, enhancing accuracy 2

   2

   .

Partnership with DataStax and Impact on Wikimedia

NVIDIA has partnered with DataStax to implement NeMo Retriever in vectorizing Wikipedia content. This collaboration has yielded impressive results:

• Processing time for 10 million Wikidata entries reduced from 30 days to under three days 1

  1

  .
• Real-time updates for hundreds of thousands of entries edited daily by thousands of contributors 1

  1

  .
• 35 times reduction in data storage volume compared to traditional approaches 3

  3

  .

Industry Adoption and Applications

Several leading companies are already adopting NeMo Retriever microservices:

• DataStax, Cohesity, Cloudera, Nutanix, SAP, VAST Data, and WEKA are integrating the technology to connect custom models to diverse data sources 1

  1

  .
• Financial services firms, including Commonwealth Bank of Australia and Capital One, are early adopters despite regulatory constraints 3

  3

  .

Future Developments

NVIDIA is working on expanding NeMo Retriever's capabilities:

• Multimodal data processing, including images, PDFs, and video 2

  2

  .
• Advanced PDF processing to understand and relate tables, graphs, charts, and images across pages 3

  3

  .

Availability and Access

Developers can access NeMo Retriever microservices through the NVIDIA API catalog or with a no-cost, 90-day NVIDIA AI Enterprise developer license 1

1

. This offering provides enterprises with a path to make their information assets AI-ready without compromising security or incurring excessive storage costs 3

3

.