Smallest.ai Closes 13 Million Dollar Series A to Scale Voice AI Across Enterprise India
On 30 July 2026, Bengaluru-based AI startup Smallest.ai closed a 13 million dollar Series A funding round led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital. The raise brings the company's total funding to more than 21 million dollars, adding to the 8 million dollar seed round it closed in 2025. Smallest.ai builds ultra-low-latency voice AI models: its Electron V2 model achieves a time-to-first-token of 53 milliseconds and generates 10 seconds of speech in 100 milliseconds. The company was founded in 2023 by Sudarshan Kamath and Akshat Mandloi, both former engineers at Robert Bosch GmbH, and has grown from a two-person operation to a startup serving more than 5,000 businesses and enterprises across India and the United States. The Series A proceeds will be directed toward model development and an enterprise sales expansion targeting the banking, financial services, insurance, healthcare, and telecom sectors in India.
The Latency Problem That Has Held Back Voice AI
Standard text-to-speech and voice AI systems built on autoregressive language model architectures generate speech token by token in sequence. This sequential generation produces a time-to-first-token of 300 to 600 milliseconds in typical production systems — a delay that is imperceptible in a podcast player but disqualifying in a real-time telephone conversation. A voice bot handling a customer service call, a healthcare scheduling application taking appointment requests, or a BFSI agent verifying a customer's identity over the phone must respond as naturally as a human agent would. At 300 milliseconds to first token, AI voice agents introduce a pause that callers recognise as unnatural and that operational managers see in post-call survey data as a satisfaction driver. Smallest.ai built its technology on a non-autoregressive architecture, which generates speech in parallel across the entire output sequence rather than sequentially. The result is the 53 millisecond time-to-first-token that its Electron V2 model delivers — roughly five to ten times faster than typical production voice AI systems and fast enough to eliminate the perceptible gap between a caller's question and the AI agent's response.
Language Coverage for India's Enterprise Market
Smallest.ai's models support more than 30 languages, with English and Hindi production-ready at the accuracy levels required for regulated sectors. For enterprises deploying voice agents in India, multilingual support with code-switching — the ability to handle a caller switching between Hindi and English mid-sentence without routing to a different model — is a practical necessity rather than a differentiator. Single-language voice AI systems require custom engineering to support this pattern; Smallest.ai's architecture handles it natively within the same model. The company supports 30-plus languages in a single model, reducing the integration complexity for enterprises that need to serve callers across multiple Indian languages in the same deployment.
From Seed to Series A: The Traction That Made the Case
Between the 2025 seed round and the July 2026 Series A, Smallest.ai grew its customer base to more than 5,000 businesses and enterprises. Two enterprise customers are publicly confirmed: RingCentral, the US-based cloud communications and contact centre platform, and Truecaller, the Swedish-Indian caller identification and spam filtering service used by hundreds of millions of users in India. The presence of both customers in the portfolio indicates that Smallest.ai's latency claims have been validated in production at commercial scale. RingCentral's use case maps to automated call flows and agent-assist applications within its contact centre product. Truecaller's use case relates to the voice intelligence and call screening features the company has been integrating into its application for the Indian market. A 5,000-plus customer base spanning SMBs and enterprises across India and the United States in fewer than three years from founding represents a growth trajectory that justified the Series A valuation and will support the enterprise sector expansion the company is planning.
The Market Smallest.ai Is Targeting With the Proceeds
The BFSI sector in India processes hundreds of millions of inbound customer service calls annually across insurance renewal, KYC verification, loan servicing, account inquiries, and fraud alerts — each a workflow where automation is economically attractive but where the latency and accuracy requirements of prior voice AI solutions disqualified them from deployment in regulated, customer-facing contexts. Healthcare scheduling and outbound appointment reminder workflows face similar requirements. Telecom customer service, one of the highest-volume call categories in India, is another deployment target. Smallest.ai is positioning its 53 millisecond time-to-first-token and 30-plus language support as the capability combination that makes enterprise-grade voice AI viable for these sectors for the first time at scale.
What This Means for Indian Software Development Teams
For Indian software agencies and product companies building customer engagement platforms, contact centre orchestration tools, or healthcare workflow automation, Smallest.ai's Series A represents a signal that sub-100-millisecond voice AI is now commercially accessible at the API level without in-house model development. Teams building on Smallest.ai's API can integrate the 53 millisecond voice generation capability into call centre software, IVR replacement systems, or outbound notification platforms without the model research and infrastructure investment that producing that capability from scratch would require. The Series A also matters from a vendor stability perspective: a startup that has raised more than 21 million dollars total with confirmed enterprise customers including RingCentral and Truecaller represents a more durable API dependency than a pre-revenue model provider. For teams building for India's domestic enterprise market, the native multilingual support covering English-Hindi code-switching is directly relevant to the use cases that have historically been the hardest to serve with off-the-shelf voice AI.
The Bottom Line
Smallest.ai, a Bengaluru-founded AI startup building ultra-low-latency voice AI models, closed a 13 million dollar Series A on 30 July 2026 led by Seligman Ventures with Sierra Ventures and 3one4 Capital participating. Founded in 2023 by Sudarshan Kamath and Akshat Mandloi, the company's Electron V2 model achieves a time-to-first-token of 53 milliseconds — roughly five to ten times faster than typical production voice AI systems — and generates 10 seconds of speech in 100 milliseconds using a non-autoregressive architecture. The company serves more than 5,000 businesses including RingCentral and Truecaller, supports 30-plus languages with native English-Hindi code-switching, and will use the Series A proceeds to expand into India's BFSI, healthcare, and telecom sectors.
Frequently Asked Questions
What does Smallest.ai build and what makes its technology distinctive?+
Smallest.ai builds ultra-low-latency voice AI models for enterprise applications, including contact centre automation, healthcare scheduling, and outbound customer communication workflows. The company's core technological distinction is its non-autoregressive architecture: while standard text-to-speech and voice AI systems built on autoregressive language model designs generate speech token by token in sequence — producing a time-to-first-token of 300 to 600 milliseconds — Smallest.ai's models generate speech in parallel across the output sequence. The result is a time-to-first-token of 53 milliseconds on its Electron V2 model and the ability to generate 10 seconds of speech in 100 milliseconds, which eliminates the perceptible delay that has historically distinguished AI voice agents from human agents in real-time telephone contexts. The models also support more than 30 languages with native English-Hindi code-switching, making them suited to India's enterprise voice AI market where multilingual caller handling is a baseline requirement.
Who are the founders of Smallest.ai and who invested in the Series A?+
Smallest.ai was founded in 2023 by Sudarshan Kamath and Akshat Mandloi, both former engineers at Robert Bosch GmbH. The company is based in Bengaluru, India, with a presence in San Francisco. The 13 million dollar Series A closed on 30 July 2026 was led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital. The round brought the company's total funding to more than 21 million dollars, following an 8 million dollar seed round in 2025 that was co-led by Sierra Ventures and 3one4 Capital. The company has grown from a two-person founding team in 2023 to a startup serving more than 5,000 businesses and enterprises across India and the United States within three years.
Which enterprise customers does Smallest.ai currently serve?+
Smallest.ai has publicly confirmed two enterprise customers: RingCentral and Truecaller. RingCentral is a US-based cloud communications and contact centre platform whose use of Smallest.ai maps to automated call flows and agent-assist applications within its contact centre product. Truecaller is the Swedish-Indian caller identification and spam filtering service used by hundreds of millions of users in India, and its use of Smallest.ai relates to the voice intelligence and call screening capabilities the company has been integrating into its India-focused application. Beyond these two named enterprise customers, Smallest.ai reports a customer base of more than 5,000 businesses and enterprises across India and the United States across its SMB and enterprise tiers.
Which sectors is Smallest.ai targeting with its Series A funding in India?+
Smallest.ai is directing its 13 million dollar Series A proceeds toward model development and an enterprise sales expansion in India targeting the BFSI sector — banking, financial services, and insurance — along with healthcare and telecom. In BFSI, the primary use cases are inbound customer service automation (insurance renewal, KYC verification, loan servicing, account inquiries, and fraud alert calls), where latency and accuracy requirements have previously disqualified available voice AI solutions. In healthcare, the target workflows include appointment scheduling, outbound reminder calls, and patient intake voice interactions. In telecom, the focus is on high-volume customer service call automation. The company was in early conversations with BFSI, healthcare, and telecom majors to deploy hundreds of voice agents across India, with the Series A capital intended to accelerate those commercial deployments.
Written by
TechPillow Team
Sharing insights on technology, product development, and the Indian tech ecosystem.