Personalized AI Language Education — with Andrew Hsu, Speak
Latent Space: The AI Engineer Podcast
2025/07/11
Personalized AI Language Education — with Andrew Hsu, Speak
Personalized AI Language Education — with Andrew Hsu, Speak
Latent Space: The AI Engineer Podcast
2025/07/11
In this episode, we explore the evolution of Speak, an AI-driven language learning platform that has grown from a niche startup into a global educational innovator. Founded in 2016, Speak leverages cutting-edge speech and language models to deliver personalized, adaptive tutoring—an approach once only possible with human instructors. The conversation with Andrew Hsu, CTO and co-founder, delves into how Speak harnessed advancements like Whisper and GPT to transform from a practice tool into a full-featured tutor.
Speak’s journey began with a vision to build the third generation of language learning tools, surpassing traditional methods like Rosetta Stone and Duolingo. Initially facing slow growth, the company found early traction in South Korea, a market with high demand for English fluency. By combining AI with deep localization, Speak achieved strong product-market fit. The arrival of advanced LLMs in 2022 allowed Speak to evolve into a comprehensive language tutor with real-time feedback and immersive roleplay. To scale globally, the company is investing in AI-generated content and developing a knowledge graph to personalize learning paths. Beyond language, Speak sees potential applications across broader educational domains. Despite AI's rapid progress, adoption remains gradual, underscoring the importance of building consumer-focused, AI-native solutions that deliver tangible value.
00:05
00:05
Andrew Hsu joins the Latent Space podcast as guest
02:38
02:38
AI language tutors will become superhuman in 5-10 years.
03:47
03:47
The company is 80-90% of the way to achieving its AI tutor vision.
13:50
13:50
Company surpasses $50 million ARR with growing B2B segment
16:30
16:30
Localization efforts impressed Korean users with cultural understanding.
22:05
22:05
90% of the core product development team is now based in San Francisco
26:21
26:21
Building a real-time voice platform for immersive language lessons.
34:40
34:40
AI enables potential to launch 100x more courses with human review for quality.
35:32
35:32
Communication ability is mostly independent of pronunciation.
47:44
47:44
Real-time API allows for interactive, lesson-like experiences but faces cost and infrastructure challenges.
57:34
57:34
Andrew reflects on using AI for transformative learning experiences.
