MemVerge.ai Intelligent Memory stores memory from past interactions between human and machines and recalls the most relevant memory to enrich the context of future queries. Working across multiple sessions, multiple agents, and multiple LLM models, Intelligent Memory captures a deep and evolving portrait of each human user to deliver more personalized AI responses and meets the security and compliance requirements needed for enterprises to deploy in their private environments. The results are accelerated task completion, better quality of output, and improved employee satisfaction.  

Utilization of scarce GPU resources is maximized with GPU orchestration. Supporting both Nvidia and AMD GPUs, the software provides enterprises with the ability to keep a watchful eye on their precious GPU resources, share them so the scarce GPU cycles are not wasted, and intelligently schedule access based on the priority of the projects. The results are lower computing costs, more AI workloads are supported, and complex optimization is available to all users because it’s automated.

GPUs can fail, and enterprises need the ability to migrate workloads. With the ability to suspend and resume jobs, workloads can surf GPUs for available capacity, hot-restart after node maintenance, and burst into another department’s GPUs during periods of peak usage. The results are lower computing cost with greater workload resilience and performance.

The SpotSurfer feature in Memory Machine Batch expands the capability of AWS Batch environments to bring more workloads onto low-cost Spot instances. The service combines checkpointing with cloud automation to gracefully handle spot terminations, making it possible for big stateful workloads to run safely on Spot instances, and allowing you to save up to 90% in compute cost.

Memory Machine X software is revolutionizing how memory is used by automatically tiering DIMM and new CXL® memory, and by allowing multiple servers to access shared memory in fabric-attached memory systems. Server Memory Expansion software ensures bandwidth or latency quality-of-service (QoS) by intelligently tiering of lower cost CXL memory. Fabric-Attached Memory software delivers 280% better Ray performance for array shuffle across 4 nodes.