ISSEN Turns a Real-Time AI Voice Into an Always-On Language Tutor
ISSEN, a Y Combinator Fall 2024 company, is betting that real-time voice AI can finally close the gap between vocabulary drills and real conversation — and has built an AI conversational tutor designed to make fluency, not streaks, the goal. Language apps have gotten very good at vocabulary drills and very bad at conversation. Most learners can conjugate verbs long before they can order coffee, hold a phone call, or navigate a new city in a second language.
Founded in 2024 and based in New York, ISSEN positions itself as “your AI realtime voice companion for language learning.” The product functions as an on-demand voice tutor that adapts lessons and conversations to each learner’s interests, learning style, and goals. Instead of moving users through a rigid curriculum, ISSEN meets them where they are — discussing topics they care about, correcting them when they stumble, and pushing them just past their comfort zone in a way human tutors have long been able to provide but most software cannot. The mission, as the company frames it, is to offer “a level of connection, effectiveness and fun that, until now, only human teachers could provide.”
The company is led by founder and CEO Mariano Sorgente, whose background spans both crypto and AI and includes time as a partner at Andreessen Horowitz. Operating thus far as a one-person show — a deliberate early-stage choice — Sorgente brings both the venture perspective on what makes language-learning companies succeed or fail at scale, and the technical pragmatism to ship a voice-first product in an area where latency, accent handling, and emotional tone all matter deeply. The bet is that ISSEN’s combination of editorial judgment (what kinds of conversations actually build fluency) and real-time AI (how to make those conversations feel alive) will differentiate the product from the crowded field of incumbents and AI-wrapper apps. The company is also supported by Joaquin Villalba Diaz, Head of Growth.
The timing is sharp. Real-time speech models have advanced dramatically over the last 18 months, finally making sub-second, expressive, and contextually aware conversation possible outside of a research lab. That capability unlocks the single most underserved use case in language learning: the conversational practice that learners crave, are afraid of, and rarely get enough of. Meanwhile, global demand for English, Spanish, Mandarin, French, and other major languages continues to grow, with hundreds of millions of adult learners who can afford a consumer subscription but have never found software that delivers genuine speaking fluency.