SC25: Pliops Accelerates GenAI Data Access and Announces Collaboration with Zilliz

At SC25, Pliops is to showcase major advancements in AI infrastructure and announce a new collaboration with Zilliz aimed at enabling affordable, large-scale Retrieval-Augmented Generation (RAG).This initiative targets multi-billion-scale vector databases at storage-level costs by combining Pliops’ LightningAI architecture with Zilliz’s Milvus vector database technology.

Enterprises increasingly rely on vector search and large-context retrieval to power GenAI applications. Yet scaling these workloads to billions of vectors often creates cost and performance barriers. The collaboration between Pliops and Zilliz directly addresses these challenges by combining the advanced vector database capabilities of Zilliz’s Milvus with the scale-out long-term memory of Pliops LightningAI. Together, these technologies enable efficient vector search, RAG, and large-scale context retrieval through hardware-accelerated KV-Cache Offload – bringing scalable, cost-effective inference within reach for enterprise AI environments.

Technical Advancements
Milvus is the leading open-source vector database built for high-performance AI workloads at massive scale. Its distributed, cloud-native architecture supports fast and accurate vector search across billions of embeddings, powering AI applications such as RAG, agentic AI, semantic search, and recommendation systems.

As part of this initiative, Pliops will introduce enhancements to Milvus, including:

• Storage APIs and tiering: Support for multi-tier storage to optimize cost and performance
• KV Mapping: Adding a key-value abstraction layer on top of file offsets for efficient caching and retrieval
• Dual-Tier Architecture
  • flash Tier (Hot): Ultra-low latency and high-density access
  • S3 Tier (Cold): Reliable, globally distributed backup for cost efficiency

These advancements build on a public RFC submitted to the Milvus community, introducing Near Compute Storage (NCS) as a shared hot storage layer between compute nodes and object storage. By combining Milvus with Pliops LightningAI, enterprises can unlock larger context windows and more efficient inference while reducing memory and infrastructure costs.

Executive Quotes
“LightningAI is designed to make AI inference scalable and affordable. Partnering with Zilliz brings the best of storage and retrieval intelligence together,” said Ido Bukspan, CEO, Pliops.

“Pliops’ LightningAI introduces a breakthrough approach to scaling GenAI inference, and integrating it with Milvus unlocks truly massive context retrieval at a fraction of the traditional cost. As the creators of Milvus, we’re committed to advancing what’s possible in vector search. This collaboration gives enterprises a clear path to run larger models, access more knowledge, and deliver faster AI experiences – all without the memory limitations that have constrained GenAI until now,” said Charles Xie, founder and CEO, Zilliz.

Availability
Technical details, including the Milvus RFC for NCS, will be published on GitHub.

GenAI Ecosystem Expansion with Viking Enterprise Solutions
At SC25, Pliops will showcase its turnkey KV-Cache offload (KV-CO) and LLM inference platforms, powered by the innovative LightningAI memory architecture. These platforms will be featured by industry leaders Viking Enterprise Solutions at booth 1107 and Giga Computing at booth 1117. By combining Pliops’ hardware-accelerated KV processing with high-performance server designs, these integrations allow organizations to run larger models, serve more users per GPU, and scale deployments across both data center and edge environments.

Pliops will demonstrate its LightningAI innovations live at SC25, Booth #4608, in St. Louis, MO November 18-20.