Gimlet Labs

About Gimlet Labs

Gimlet Labs is an applied research lab developing an inference cloud platform designed to run and manage complex AI agent workloads. The company's core technology orchestrates AI tasks across diverse and heterogeneous hardware systems, including GPUs and other specialized accelerators, to improve performance and efficiency. Its software divides AI workloads and routes different parts to the most suitable hardware, reporting significant inference speed enhancements for the same cost and power usage. The Gimlet Cloud product provides serverless inference, allowing developers to build and scale multi-agent systems that integrate various models, data sources, and custom logic, while the platform manages scheduling and optimization. The company was founded in 2023 by CEO Zain Asgar, Michelle Nguyen (Head of Engineering), Omid Azizi (Head of Hardware Platforms), Natalie Serrino (leading KForge), and James Bartlett (System Architect). The founding team had previously worked together at Pixie Labs, an open-source observability company acquired by New Relic. Gimlet Labs originated from a Stanford University research project, driven by the founders' belief that AI inference would become a primary infrastructure challenge. The company asserts that as agentic AI applications generate substantially more tokens than traditional models, the computational demand requires a new approach to hardware orchestration. Gimlet Labs offers its platform as a cloud service and also allows customers to deploy the software within their own data centers. In addition to its main cloud platform, the company is developing a toolkit called kforge, which uses AI agents to automatically generate optimized computing kernels from PyTorch for various hardware platforms like CUDA, ROCm, and Metal. The firm emerged from stealth in October 2025 and reported a tripling of its customer base by March 2026. Total funding reached $92 million after an $80 million Series A round in March 2026, led by Menlo Ventures. A subsequent $300 million round led by Andreessen Horowitz in September 2026 brought the company's valuation to $3 billion. Keywords: AI inference cloud, serverless inference, AI agents, heterogeneous hardware, workload orchestration, automatic kernel generation, multi-silicon inference, kforge, AI infrastructure, high performance AI systems, compute optimization, AI pipelines, datacenter optimization, GPU kernel generation, Zain Asgar, Michelle Nguyen, Omid Azizi, Natalie Serrino, James Bartlett, Stanford research, edge AI, AI workload management, PyTorch optimization, AI systems