Today we are excited to announce our investment in Hearvana, the acoustic layer for personalized intelligence. The company raised a $6 million round with participation from Point72, the AI2 Incubator, Amazon Alexa, and SBI Investments.

Audio remains one of the most important and least developed interfaces for AI. Most devices treat sound as a simple input channel and struggle in real-world conditions. Noise cancellation suppresses everything. Voice assistants fail when the environment gets loud. Smart glasses and ambient agents depend on accurate context but cannot reliably detect or separate the conversations happening around the user. The gap between what people expect from natural audio interaction and what current hardware can deliver grows more obvious each year.

Hearvana’s technology creates programmable sound bubbles that allow users to focus on selected voices and filter out surrounding noise, which goes beyond noise cancellation. The company demoed real-time neural networks that run on embedded CPUs and separate voices based on distance, phase, and multi-channel audio features. The system processes 8 millisecond audio chunks with total latency under 10 milliseconds, which preserves lip sync, supports spatial alignment, and enables on-device AI that reacts quickly and consistently.

The result, if successful, is an acoustic foundation that can support the next wave of human-computer interaction. Hearvana plans to develop an SDK that integrates into headphones, hearables, hearing aids, and AR glasses. This enables, for instance, recording of only the voice of a conversation partner in a noisy restaurant. It can transcribe a multi-party discussion and highlight missed dialogue. It can act as an in-ear assistant during meetings. It can also power real-time selective listening for people with hearing loss. These capabilities create value across consumer audio, enterprise communication, health, accessibility, and emerging spatial computing platforms.

The founding team is one of the strongest we have seen in applied audio AI. CEO Shyam Gollakota spent more than 12 years as a professor at the University of Washington where he invented breakthrough acoustic systems. He previously co-founded Sound Life Sciences, which secured FDA 510(k) clearance for its respiratory monitoring app before being acquired by Google. Co-founder Malek Itani spent four years as a researcher at the University of Washington and authored the first open-source speech AI paper for wireless earbuds. Co-founder Tuochao Chen developed sound AI at Meta and Google and holds a PhD in computer science from the University of Washington. Three of the five authors of the Nature Electronics paper on sound bubbles are part of the founding team.

Hearables represent a global $40 billion category with strong growth. Hearing aids generate $8 billion annually with customers who pay premiums for enhanced clarity. More than 1.5 billion people worldwide experience some level of hearing loss. AI assistants and smart glasses depend on accurate real-world audio capture and fail without it.

Competition is increasing but remains fragmented. Traditional players like Phonak and newer entrants like Google are embedding AI accelerators into earbuds and hearing aids. Syntiant has raised more than $121 million to run low-power audio models on custom silicon. Hearvana differentiates by offering a hardware-agnostic, edge-optimized model that already runs at roughly 4.7 GOPS on commodity boards like Raspberry Pi. This means customers do not need custom chips to benefit from advanced selective listening. The academic depth of the team gives Hearvana a lead in model design. The early industry validation signals that their approach translates beyond the lab. The company also focuses on real-world constraints such as power budgets, latency limits, and multi-channel alignment, which positions the platform well for both consumer and medical-grade deployments.

We invested because Hearvana is building the acoustic intelligence layer that modern AI systems need. The company combines foundational research, tight engineering, and practical efficiency on constrained hardware.

We are proud to back the Hearvana team from day one.