An Uzbekistan-born founder, a Y Combinator alumnus, has launched Speko on Product Hunt. Speko is a startup that provides infrastructure for voice AI services. The company, founded in San Francisco in 2026, offers developers a unified API to select the most effective speech recognition models, large language models, and speech synthesis models depending on the language and specific task.

Speko constantly compares available models by recognition accuracy, latency, response generation speed, and cost per minute of operation. The results are published on the open website benchmarks.speko.ai. If the performance of one provider deteriorates, the platform automatically reroutes requests. In case of a service failure, it can switch to a backup provider in less than 500 ms.

The project was founded by Beknazar Abdikamalov, who previously worked as co-founder and CTO of Hupo. In four years, the company raised $14 million and created voice AI solutions for projects like Morgan Stanley, Prudential, HSBC, and Grab. It was the need to manually compare speech models for different markets that prompted Abdikamalov to create Speko.

Before Hupo, the entrepreneur worked as a software engineer at Amazon and was among the first engineers at Tridge. He got into Y Combinator on his fourth attempt, becoming the first solo-founder from Uzbekistan accepted into the accelerator. Currently, the Speko team consists of four engineers from Central Asia. According to Abdikamalov, the project aims to show young developers from Nukus and Tashkent that to build a tech company in Silicon Valley, you don't necessarily have to be born there.