Luminal Raises $5.3 Million To Build A Better Gpu Code Framework

Three years ago, Luminal co-founder Joe Fioti was moving connected spot creation astatine Intel erstwhile he came to a realization. While he was moving connected making nan champion chips he could, nan much important bottleneck was successful software.

“You tin make nan champion hardware connected earth, but if it’s difficult for developers to use, they’re conscionable not going to usage it,” he told me.

Now, he’s started a institution that focuses wholly connected that problem. On Monday, Luminal announced $5.3 cardinal successful seed funding, successful a information led by Felicis Ventures pinch angel finance from Paul Graham, Guillermo Rauch, and Ben Porterfield. 

Fioti’s co-founders, Jake Stevens and Matthew Gunton, travel from Apple and Amazon, respectively, and nan institution was portion of Y Combinator’s Summer 2025 batch.

Luminal’s halfway business is simple: nan institution sells compute, conscionable for illustration neo-cloud companies for illustration Coreweave aliases Lambda Labs. But wherever those companies attraction connected GPUs, Luminal has focused connected optimization techniques that fto nan institution compression much compute retired of nan infrastructure it has. In particular, nan institution focuses connected optimizing nan compiler that sits betwixt written codification and nan GPU hardware — nan aforesaid developer systems that caused Fioti truthful galore headaches successful his erstwhile job.

At nan moment, nan industry’s starring compiler is Nvidia’s CUDA strategy — an underrated constituent successful nan company’s runaway success. But galore elements of CUDA are open-source, and Luminal is betting that, pinch galore successful nan manufacture still scrambling for GPUs, location will beryllium a batch of worth to beryllium gained successful building retired nan remainder of nan stack.

It’s portion of a increasing cohort of inference-optimization startups, which person grown much valuable arsenic companies look for faster and cheaper ways to tally their models. Inference providers for illustration Baseten and Together AI person agelong specialized successful optimization, and smaller companies for illustration Tensormesh and Clarifai are now popping up to attraction connected much circumstantial method tricks.

Luminal and different members of nan cohort will look stiff title from optimization teams astatine awesome labs, which person nan use of optimizing for a azygous family of models. Working for clients, Luminal has to accommodate to immoderate exemplary comes their way. But moreover pinch nan consequence of being out-gunned by nan hyperscalers, Fioti says nan marketplace is increasing accelerated capable that he’s not worried.

“It is ever going to beryllium imaginable to walk six months manus tuning a exemplary architecture connected a fixed hardware, and you’re astir apt going to hit immoderate sorts of, immoderate benignant of compiler performance,” Fioti says. “But our large stake is that thing short of that, nan all-purpose usage lawsuit is still very economically valuable.”