Gimlet Labs, Inc.

This post is a little late but I am thrilled to announce that we have invested in Gimlet Labs, Inc.!! I am really excited to work with my friend Zain Asgar again, and welcoming co-founders Natalie Serrino, Michelle Nguyen and Omid Azizi to the Samsung Ventures family. Gimlet Labs, Inc. is building what may be one of the most compelling “multi-silicon” inference clouds in the market today.🚀 We are doubling down on AI infrastructure at Samsung Ventures and Gimlet’s approach to decoupling agentic workloads from specific hardware and orchestrating them across heterogeneous accelerators (GPUs and SRAM‑centric silicon) really resonates with how we see the future of efficient AI at scale. It’s already impressive to see the team translate applied research into a platform that is already delivering multi‑x gains in inference efficiency for demanding, agentic workloads. Also shout out to our co‑investors Menlo Ventures, Eclipse, Factory, Prosperity7 Ventures, Triatomic Capital, and an outstanding group of operators and technical angels who have been early champions of Gimlet’s vision.🤝 #AI #DisaggregatedInference #AgenticAI #Datacenter #SamsungVentures https://lnkd.in/gMgZRSdC

Gimlet is honored to be named in The Compute 100! The enormous demand for agentic inference, and its unique needs from hardware, are driving revolutionary innovation in compute. Gimlet was built on the thesis is that no single chip architecture can do the job alone, and heterogeneous hardware is needed to run agents at the scale we need. This is a great list of companies building at the heart of the industry’s need, including partners of ours. Check it out below!

Check out Gimlet's appearance on the Chipstrat podcast with Austin Lyons - Natalie Serrino and Beltir G. Çağlar Dayanik discuss the following: - The unique hardware needs of agentic inference - What that means for the future of datacenters - What we're building at Gimlet to unlock better performance, capacity, and capabilities for these workloads

We're #hiring a new Executive Business Partner in San Francisco, California. Apply today or share this post with your network.

Incredibly excited to collaborate with Primary Venture Partners next Wednesday to host The Future of Compute event, featuring leaders from Etched, Gimlet Labs, Inc., Substrate, Lightmatter, and Together AI! Apply to attend here: https://luma.com/y0dix0fp

We are thrilled to welcome Elliot as our Head of Talent Acquisition. We're hiring across the stack, so if you're interested in building the future of AI infrastructure with us, connect with him below!

If you're at EuroSys, come check out Gimlet's session at InSyDe by Taras Sereda on Gimlet's work in kernel generation. Link: https://lnkd.in/gH9jkvUb Time: Monday, April 27 at 9:45am

In case you missed it: here's the recording from Chris Fregly and Antje Barth's AI Performance Engineering meetup, where Natalie Serrino breaks down the following: - Why agentic inference requires heterogeneous systems - The technical problems that need to be solved to deploy agentic workloads across heterogeneous hardware - The types of results and outcomes these systems can deliver Link here: https://lnkd.in/gzzcZxGw

Tomorrow 4/20 at 9am Pacific, Natalie from Gimlet will be joining Chris Fregly and Antje Barth for their AI Performance Engineering meetup, discussing why hardware heterogeneity is needed for high performance inference. Sign up below! Link: https://lnkd.in/gUtSKVda

This is the best blog post on LLM inference I've seen this year. Gimlet Labs, Inc. achieved 10x latency and >1400 tokens/sec by moving speculative decode onto two 2GB SRAM/chip d-Matrix Corsairs, a small cost on top of a standard GPU setup on gpt-oss-120b. This performance at this price is insane. Source: https://lnkd.in/gApq76BU