Smallest AI: Building the Voice Infrastructure for AI Agents
A voice AI research and infrastructure company building compact, specialized models for real-time speech recognition, text-to-speech, speech-to-speech and conversational AI agents at enterprise scale.
Visit smallest.ai ↗Basic Information
| Field | Detail |
|---|---|
Company Name |
Smallest AI |
Category |
AI Infrastructure |
Subcategory |
Voice AI, Speech Models & AI Agents |
Founded |
2023 |
Headquarters |
San Francisco, California, USA |
Stage |
Series A |
Total Funding |
More than $21 million |
Latest Funding |
$13 million · Series A · July 2026 |
Founders |
Sudarshan Kamath & Akshat Mandloi |
Leadership |
Sudarshan Kamath, CEO · Akshat Mandloi, CTO |
Key Investors |
Seligman Ventures, Sierra Ventures, 3one4 Capital & Better Capital |
Website |
smallest.ai ↗ |
What Smallest AI Does
Smallest AI builds voice models and infrastructure for real-time conversational systems. Its stack covers speech recognition, text-to-speech, speech-to-speech and compact language models designed for low-latency enterprise voice applications.
Key Capabilities
- Real-time speech-to-text with multilingual support
- Low-latency text-to-speech for conversational agents
- Speech-to-speech models designed for natural conversations
- Compact language models optimized for voice workloads
- Visual agent builder, telephony, workflows and analytics
Why It Matters
Voice agents have stricter latency requirements than text interfaces. Smallest AI is taking a specialized-model approach, using smaller systems optimized for speech so agents can listen, reason and respond with less delay.
Smallest AI at Scale
Millions
Call Minutes / Month
Smallest AI says its technology handles millions of call minutes every month across production voice applications.
38
STT Languages
The company’s Pulse speech-to-text product currently supports transcription across 38 languages.
15+
TTS Languages
Its Lightning text-to-speech system supports more than 15 languages with approximately 100 ms latency, according to the company.
Enterprise Momentum
Voice Agent Platform
Smallest AI combines voice models, telephony, knowledge, workflows and analytics so enterprises can build and deploy production voice agents without assembling the entire stack themselves.
Enterprise Customers
Customers cited by the company and recent coverage include RingCentral and Truecaller, while the platform is designed for customer support, sales, healthcare, collections and other high-volume voice workflows.
Enterprise Security
Smallest AI lists SOC 2 Type 2, ISO 27001, GDPR and HIPAA compliance and supports private-cloud and on-premise deployment for higher-trust environments.
Voice Agent Infrastructure
Pulse STT
A production speech-to-text model designed for high accuracy, low latency, multilingual transcription and speaker and emotion detection.
Lightning TTS
A low-latency text-to-speech model designed to generate natural speech for real-time agents, with support for more than 15 languages.
Hydra
Smallest AI’s speech-to-speech architecture is designed to process conversational audio continuously rather than forcing every turn through a sequential speech pipeline.
Funding Journey
2023
Company Founded
Smallest AI
Founded by Sudarshan Kamath and Akshat Mandloi
2025
Early Funding
$8M+
Early institutional capital before Series A
July 2026
Series A
$13M
Led by Seligman Ventures
July 2026
Total Funding
$21M+
Company-reported cumulative funding