In Uzbekistan, English proficiency remains one of the essential skills for admission to foreign universities and for career development. IELTS preparation is particularly relevant, and the Speaking section frequently becomes a major obstacle for students. Training centers often face resource limitations: teachers cannot dedicate sufficient time to each learner, while individual lessons remain comparatively expensive.

 

Speaklish — an AI-powered service for IELTS Speaking training — offers a solution to this challenge and has attracted more than 50,000 users within a short period.

 

From idea to mission

 

Speaklish was founded in 2023 by Abdul Sharopov and Bobir Mardonov. At the initial stage, the team consisted only of the two founders, and the project developed without substantial investment. They choseto focus on the Speaking section because it poses the greatest difficulties for students and requires individualized practice.

 

The idea was inspired by an observation of the local market: educational centers struggle to provide personalized speaking practice, leaving students to face their uncertainties alone.

 

“We saw that students were getting far less practice than they needed, and teachers simply could not keep up. AI bridges this gap,” the team notes.

 

Speaklish was launched both as a mobile application and a Telegram bot. The bot quickly became the primary growth driver: due to convenient onboarding and the audience’s familiarity with Telegram, the first 10,000 users joined within two months, without significant marketing expenses.

 

Product and technology

 

Speaklish evaluates responses according to official IELTS criteria: fluency, grammar, lexical resource, and pronunciation. Additionally, the platfrom provides AI-driven metrics, including the number of filler words, pause length, and common mistakes, making the feedback more actionable.

 

Its reliability is supported by practice: Speaklish scores align with certified examiner assessments within a 0.5 band range, based on more than 500 comparisons. Special attention is given to accent neutrality: the model is trained on more than 10,000 voice samples from Central Asia and undergoes regular bias audits — evaluations that ensure fair scoring regardless of a student’s accent or origin.  

 

Experience demonstrates that fluency depends not only on grammar but also on confidence. Many students experience stress during the exam. For this reason, Speaklish introduced psychological support elements and gamified features to reduce anxiety.

 

Importantly, the service does not store users’ voice data. All audio files remain on Telegram’s servers, and users may delete their recordings at any time.

 

Market and users

 

Speaklish is currently used by more than 50,000 individual learners — a milestone achieved in less thantwo years. The majority of users are between 18 and 30 years old and dedicate an average of 30–40 minutes per day to practice. The user base has expanded beyond Uzbekistan: organic growth is observedin Nigeria, India, and Iran, driven by social media virality and word-of-mouth recommendations. Nevertheless, the team maintains strategic focus on Kazakhstan and Saudi Arabia, where monetization and scaling potential are considerably higher.

 

Speaklish follows a flexible freemium model: the core service is free for students, while detailed feedback on a response costs only $1. For schools and training centers, a separate subscription is available — $2 per student per month. This approach makes the product accessible to a broad audience while generating steady revenue through the B2B segment. A significant portion of revenue now comes from educational organizations and online platforms that have integrated Speaklish into their teaching processes. This enables consistent income growth, strong conversion among partners, and allows the team to scale confidently without reliance on external funding.

 

Beyond individual users, the team actively develops white-label solutions for educational institutions. One example is the Daryn school, which implemented a customized version of the Speaklish voice assistant (@DOspeaking_bot). Several other integrations with schools and language platforms are currently underway.

 

Team and approach

 

The core development of Speaklish is led by CTO Bobir Mardonov, previously a data engineer at Korzinka. Under his leadership works a compact yet highly efficient team, including specialists in machine learning, backend engineering, and mobile development. Contract-based engineers and designers join during specific sprints. The team emphasizes that a well-coordinated and deeply engaged group often outperforms larger distributed teams — thanks to strong product understanding, close collaboration, and flexibility. This enables the startup to rapidly implement new features and respond to user needs without excessive bureaucracy.

 

Speaklish combines pedagogical expertise with technological solutions. The team works closely with IELTS instructors and academic experts on a continuous basis: they refine methodology, identify common student errors, and test new task formats. Developers then incoraporate this feedback into the app’s logic, ensuring that AI delivers clear, supportive, and approachable feedback. Thus, learners receive not only technical assessments but also friendly guidance, for example: “Great job! Try reducing the ‘uh’ fillers — here’s a tip.”

 

Recently, Speaklish launched an interactive web application that allows learners to practice directly within Telegram, without installation. The service also became available on Hambi (Beeline Uzbekistan), opening a new and high-growth user acquisition channel.

 

Investment and future plans

 

In 2024, Speaklish raised $15,000 from the Werise accelerator and an additional $50,000 through the President Tech Awards. These investments played a crucial role in scaling the project. The funds were allocated to enhancing the AI speech-recognition and answer-analysis models, as well as expanding the functionality of the Telegram bot, which remains one of the most popular usage channels.

 

Part of the grant funding was invested in marketing campaigns across Central Asia to strengthen brand awareness and attract new learners from the region.

 

Advice from the founder to early-stage EdTech startups

 

1. Address one specific problem deeply rather than trying to solve everything at once.

2. Leverage familiar platforms (Telegram, WhatsApp) for rapid market entry.

3. Launch an MVP as early as possible to validate demand before scaling.

 

Where to follow Speaklish

 

• Website: speaklish.uz

• Instagram: @speaklish.uz

• Telegram: t.me/speaklish_uz