Shunya Labs
Category: Voice / Speech AI
Research-driven voice AI company building CPU-first foundation models for Indian and multilingual code-switched speech. Shunya Labs was founded in 2025. The company is led by Ritu Mehrotra. Based in Gurugram, India. Team size: 11-50. Total funding raised: $5.5M. Latest round: Seed.
- Founded
- 2025
- Headquarters
- Gurugram, India
- Team size
- 11-50
- Total funding
- $5.5M
Value proposition
Foundation models for speech-to-text, text-to-speech, and voice agents trained from scratch on real-world Indian audio (200+ languages, 40+ Indian languages), optimized to run on CPUs for up to 20x lower deployment cost than GPU-based systems, with sub-100ms latency and on-prem/air-gapped deployment for regulated industries.
Products and solutions
Zero STT (speech-to-text, 3.10% WER on OpenASR leaderboard), Zero STT Indic, Zero STT Codeswitch (Hinglish/code-mixed speech), Zero STT Med (medical/clinical transcription), Zero TTS & Zero TTS Indic (text-to-speech), Zero Tiny ONNX (on-device models), Voice Agents platform, Real-Time Translation, Edge SLU, Custom SLMs (domain-specific small language models).
Unique value
First-principles, CPU-first architecture trained natively on Indian acoustic data (accents, dialects, code-switching) rather than English models retrofitted for India; best-in-class WER (3.10%), HIPAA/SOC2 Type II/ISO 27001 compliant, on-prem/air-gapped deployment.
Target customer
Enterprises and public-sector organizations in healthcare, contact centers/BFSI, fintech, government services, and media & entertainment — primarily in India, expanding internationally.
Industries served
Healthcare, Contact Centers/BFSI, Media & Entertainment, Government/Public Sector, Fintech
Technology advantage
Proprietary training methodology and model architecture trained ground-up on millions of hours of real Indian speech; unified tokenization for mixed-language (Hindi/English) transcription without intermediate translation layers; CPU-first inference reducing infra cost/dependency on GPUs.
How they differentiate
Focus on deep, first-principles research vs. short-term marketing; ground-up Indian-language training data rather than adapting global English-first models; CPU-first architecture vs GPU-dependent competitors; enterprise-grade privacy/compliance (on-prem, air-gapped).
Main competitors
Sarvam AI, Gnani.ai, Yellow.ai
Key partnerships
Nasscom (co-launched Vāķ, an open-weight real-time translation/voice AI platform spanning 55 Indian languages, unveiled at the India AI Impact Summit, Feb 2026), Open-source releases on Hugging Face, spun out of United We Care (AI mental health platform), enterprise/public-sector customers in India and abroad (unnamed).
Major milestones
Launched Zero STT (record-low WER), launched Vāķ open-weight real-time translation model spanning 55 Indian languages / 2,970 translation pairs, in partnership with Nasscom, at India AI Impact Summit (Feb 2026), launched Zero Codeswitch foundation model for Hinglish/code-mixed Indian speech (Jan 2026), launched end-to-end voice platform for dubbing/translation/localization for media & entertainment (Apr 2026), achieved HIPAA, SOC 2 Type II, ISO 27001 compliance.
Growth metrics
Achieved 3.10% Word Error Rate on OpenASR leaderboard (48% improvement over next-best competitor); plans to raise a Series A within 3-4 months of July 2026 article; prioritizing profitability.
Market positioning
Domestic (India) full-stack voice AI infrastructure provider positioning itself as sovereign/indigenous alternative to global voice AI platforms, benchmarked against global competitors.
Geographic focus
India-first, expanding to international markets
Patents and IP
12 patents, 18+ peer-reviewed publications (per CTO Sourav Bandyopadhyay's bio); models open-sourced on Hugging Face (org: shunyalabs).
About Ritu Mehrotra
2X founder — founded United We Care (AI-powered mental health platform), Shunya Labs is its deeptech voice AI spinout; previously led growth/operations at Booking.com, Zomato, Mahindra. MBA, IIFT Delhi; also attended Harvard Business School executive program. Cancer survivor and mental-health advocate.
Latest news about Shunya Labs
- Shunya Labs has launched Zero Codeswitch, a speech foundation model achieving a 3.10% Word Error Rate, a 48% improvement over top competitors. It processes code-mixed Indian speech natively, removing
- Shunya Labs launched Zero Codeswitch, a foundation model achieving a 3.10% Word Error Rate on the OpenASR leaderboard, a 48% improvement over competitors. It natively processes code-mixed speech like
More Voice / Speech AI companies
Official website: https://www.shunyalabs.ai
