Tensormesh will exhibit at The AI Conference 2026, Sept. 30–Oct. 1, at Pier 48 in San Francisco and will showcase KV caching solutions aimed at reducing redundant GPU compute. Attendees can visit Booth #214 to learn about KV cache offloading, inference cost reduction, and techniques to make GPU memory go further at scale. Junchen Jiang, Tensormesh co-founder and CEO, will be on site to speak about KV caching and AI inference economics and to present the company’s latest innovations.
KV caching and fleet-level inference
As inference deployments move from single servers to GPU fleets, identical prompts and context are often recomputed across nodes with no straightforward mechanism to share cached state. Tensormesh extends its KV caching expertise to the fleet level through the Tensormesh Platform, addressing cache reuse across nodes to cut inference costs.
Junchen Jiang, who is a University of Chicago faculty member and co-creator of the open source LMCache project, will meet with builders, researchers, and infrastructure leaders at the event to discuss KV cache reuse and what’s next for Tensormesh. Conference attendees are invited to stop by Booth #214 throughout the show for walk-throughs and conversations with the team.
Tensormesh describes itself as a leader in smart AI-native data management and caching-accelerated inference optimization for enterprise AI. The company was founded by faculty, Ph.D. researchers and alumni from the University of Chicago, UC Berkeley, and Carnegie Mellon, and is led by Junchen Jiang. Tensormesh has raised $28.5 million in total funding and is backed by Valley Capital Partners, NVentures, AMD Ventures, CoreWeave, Laude Ventures and others.
Members of the media interested in scheduling a briefing or demo with Junchen Jiang may contact PRforTensormesh@bospar.com. General media inquiries can be directed to Laura Kerl at press@tensormesh.ai.
