This project stemmed from a partnership with Sana Labs, a Sweden-based AI startup focused on education. My role involved reviewing and testing various speech recognition APIs to assess their potential for language learning applications.
The objective was to develop a solution that would enable students to practice speaking in a controlled environment, where they could receive feedback and improve their pronunciation without external pressure. Beyond evaluating APIs, the project required addressing challenges related to speech variability, accent recognition, and the effectiveness of automated feedback.
Rather than a purely technical experiment, this was an exploration of how AI-driven speech recognition could support language learners by providing structured yet flexible speaking practice.
My role was to lead the project through the product lifecycle, from initial concept and discovery to delivering a proof of concept that would guide the development of a minimum viable product. I managed both the budget and the project itself while also driving research, workshops, and design efforts.
Throughout the project, I collaborated with commercial product managers from the adult English language learning (ELT) portfolio and sales development teams in Japan. I developed the service concept, facilitated ideation workshops, oversaw development work, and contributed to coding, content creation, and learning design.
The market challenge was first identified in Japan, where students have limited opportunities to practice speaking, and teachers often struggle to provide actionable feedback—partly due to their own language proficiency. This gap has long hindered English language learning, but recent educational reforms are set to reshape the landscape.
As part of a nationwide overhaul, Japan has introduced high-stakes university entrance exams that assess all four language skills: reading, writing, listening, and speaking. Previously, speaking was not formally tested, leaving students unprepared for real-world communication. With this change, students must now take an internationally recognized English exam approved by the Ministry of Education. Additionally, Tokyo has already incorporated English-speaking tests into high school entrance exams, signaling a broader push for improved language proficiency.
This shift creates both a challenge and a market opportunity. Schools and students now face an urgent need for scalable solutions that provide effective speaking practice and reliable feedback. The demand for technology-driven language learning tools is growing, offering new possibilities to bridge the gap between classroom instruction and real-world communication skills.
The research found that students in other parts of the world (Turkey, Lat Am, Europe, US) experience similar problems related to speaking practice and formative assessment of their speaking skills. Teachers expect the solution to work outside of the classroom yet expect some links to the lessons they teach.
This initiative focused on providing a solution that would create a safe environment conducive to practicing speaking English via mobile devices. It would allow the students to interact and receive actionable feedback related to their performance and engage in activities designed to help them improve, available directly in the solution. This needs to be achieved without involving teachers and create multiple opportunities for learners to speak on a continuous basis.
An interface that requires learners to speak and only speak. Learners are not asked to type or put up with confusing instructions.
Introducing an adaptive English speaking course that focuses on building learners' accuracy and confidence in communication through authentic conversational patterns.
With its spaced repetition module, learning hints and tips ChatterBerry is ideal for self-study as a speaking practice companion to your English language course.
Chatterberry creates a safe and moderated learners' community to provide a sense of real purpose through communication with others. Relax, listen and respond.