How Luminal is Delivering AI Inference at the Speed of Light

Luminal, compiling AI models to give you the fastest, highest throughput inference cloud in the world, led by Joe Fioti, Matthew Gunton, and Jake Stevens, has secured a $5.3 million seed round led by Felicis Ventures, with incredible angels like Paul Graham, Guillermo Rauch, and many more.

Compute demand is soaring, and chips are advancing faster than ever. But software isn’t keeping pace, leaving large portions of modern accelerators underutilized. Reaching peak performance is becoming harder with each new generation, widening the gap between hardware potential and real-world efficiency.

Luminal is building a world where full hardware utilization is as simple as running luminal.deploy(). Teams can focus on product, not CUDA quirks or complex inference stacks.

By combining a high-performance compiler with an integrated inference cloud, we eliminate the software bottleneck. Large-scale kernel search unlocks speed-of-light performance across GPUs and ASICs, and a tightly integrated cloud makes that power effortless to use.

Luminal has been open source from day one, driven by strong community adoption. To deliver true speed-of-light inference for all, the core compiler must be built in the open, so developers can use it, adapt it, and run it on their own hardware.

Accelerated computing is too complex for any one company to solve alone. AI engineers who want to eliminate 90% of today’s complexity are invited to build with us.