Gimlet Labs

Gimlet Labs is an applied research lab working on AI infrastructure for what it calls the agentic era. Its central proposition is a heterogeneous neocloud - the first of its kind, the company claims - that decouples AI workloads from the underlying hardware and orchestrates each component onto the best-fitting machine. The stated aim is to make AI workloads ten times more efficient by expanding usable compute and improving orchestration, with systems designed to scale to gigawatt-class AI datacenters.

Two products sit at the core of its work. Gimlet Cloud is a serverless inference platform for AI agents. kforge is an autonomous kernel generation system that uses multi-agent architectures to explore alternative designs and identify the fastest kernels automatically. The technical range spans heterogeneous compute orchestration, serverless inference, kernel generation and optimization, workload partitioning and the raising of AI accelerator utilisation.

The platform supports NVIDIA, AMD, Intel, ARM, Cerebras and d-Matrix hardware. Gimlet Labs works with foundation model labs, hyperscalers and AI-native companies on production workloads. It was founded by Zain Asgar, an adjunct professor at Stanford, whose co-founders include Michelle Nguyen, Omid Azizi and Natalie Serrino.