Startup founded by Uzbek entrepreneur featured on Product Hunt
Speko, a startup launched in Silicon Valley by Uzbek entrepreneur Beknazar Abdikamalov, was highlighted on Product Hunt, a top platform for launching tech products worldwide. The project tackles a major challenge in the expanding global Voice AI market by helping developers identify the best solutions for specific languages and tasks.
Today, Voice AI technologies – systems capable of recognizing human speech, processing it using artificial intelligence, and engaging in meaningful real-time voice conversations – are developing at a rapid pace. However, engineers worldwide consistently face a major technical challenge. Each such system consists of three distinct components: speech recognition (Speech-to-Text), a large language model (LLM) for generating responses, and speech synthesis (Text-to-Speech). No provider can deliver all three components equally well across every language. A model that performs perfectly in English may make critical errors in Uzbek or Spanish, suffer from response delays, or be prohibitively expensive.
The Speko platform solves this issue. It creates a unified Voice AI infrastructure, so developers can avoid relying on a single provider. Using a single API, it seamlessly integrates solutions from various providers, selects the best combination for a specific language and use case, and directs requests to the most effective technology.
Models are compared using strict metrics, including Word Error Rate, latency, time to first token, and cost per minute of operation. Results from regular tests are published openly, providing real performance figures rather than promotional claims.
In the Voice AI sector, market leadership shifts constantly. Speko’s key advantage is dynamic routing. If a previously leading solution begins to lag behind competitors, the system automatically reroutes requests without any code changes. If one provider fails, Speko switches to a backup option in less than 500 milliseconds.
“No single provider can be equally effective across all areas, and the quality of AI-generated speech depends directly on the selected language”, says Beknazar Abdikamalov. “We route each request to the technology stack that delivers the best performance at that moment”.
The idea for the project grew out of practical experience. For four years, Beknazar served as Co-Founder and Chief Technology Officer (CTO) of Hupo, a company that raised $14 million. While developing Voice AI solutions for major companies such as Morgan Stanley, Prudential, HSBC, and Grab, the team manually tested models for each new market and language. The need to automate this manual process ultimately led to the concept behind Speko.
Earlier, B. Abdikamalov worked as a software engineer at Amazon and Tridge. On his fourth attempt, he was accepted into the renowned Y Combinator startup accelerator in San Francisco, becoming the first solo founder from Uzbekistan admitted to the prestigious program.
Founded in San Francisco in 2026, Speko has made a significant move into the global market with its official Product Hunt launch. In the competitive field of artificial intelligence, the Uzbek entrepreneur aims not just to find an overall “best” AI model, but to discover the most effective solution tailored for each language and task.
Muhayyo Toshqorayeva, UzA
