Sarvam AI: Founders, Funding, Valuation & Reviews

India's homegrown AI startup Sarvam AI has recently grabbed headlines by outperforming global giants on local benchmarks. According to reports, Sarvam AI's new OCR model Sarvam Vision "beats the likes of Gemini and ChatGPT in reading documents in Indian languages"^[1]. In benchmark tests, Vision scored an impressive 84.3% accuracy on the olmOCR-Bench (a multilingual OCR test), higher than Google's Gemini 3 Pro and other leading models. Its document-understanding accuracy (OmniDocBench v1.5) was even higher at 93.3%. These results have attracted international attention: tech journalists note that Sarvam Vision's performance on India-specific tasks is "world-class", and many users on social media are impressed. In short, Sarvam AI beats Google Gemini and ChatGPT on key India-focused OCR tasks, a milestone for Indian AI.

Sarvam AI isn't just about OCR. Its voice AI model Bulbul V3 delivers natural, expressive speech synthesis in multiple Indian languages. Bulbul V3 currently supports 35+ distinct voices across 11 Indian languages (and plans for 22 total). In testing, Bulbul has shown lower error rates than global speech systems on telephony-grade audio and code-mixed text. In other words, Sarvam AI's tools are purpose-built for India's linguistic mix and infrastructure. The startup calls its approach "sovereign AI", meaning AI built locally for India's needs. This means Sarvam AI models are trained on diverse Indian content (22 official languages, mixed scripts, handwritten forms).

Sarvam AI's sudden success has changed perceptions. Skeptics who once doubted the focus on small Indic models are now praising the startup. Tech commentator Deedy Das admitted "I was wrong about Sarvam" and called its models "the best text-to-speech, speech-to-text, and OCR models for Indic languages". Users on X (Twitter) have responded enthusiastically ("Oh man wow", one user wrote after trying Vision). Even AI entrepreneurs like Pratik Desai of KissanAI say Bulbul is now his "go-to tts model for Indic use cases". The message is clear: Sarvam AI has proven that a Bangalore startup can compete with tech giants on real-world Indian tasks, and the world is taking notice.

What is Sarvam AI?

Sarvam AI is a Bengaluru-based AI startup founded in August 2023. Its mission is to build "foundational AI components" tailored to India's unique needs. Instead of general-purpose chatbots, Sarvam builds compact, efficient models designed for India's languages and use cases. The company focuses on document intelligence and speech technology. Its flagship vision model – Sarvam Vision – is an OCR/document understanding system trained on high-quality datasets of Indian text (22 official languages, scanned documents, newspapers, historic texts, etc.). Vision excels at reading and extracting data from real-world Indian documents (even handwritten or poorly formatted), as reflected in its benchmark scores.

In addition to Vision, Sarvam AI develops a suite of language and voice tools. The Bulbul V3 model is a text-to-speech engine that generates natural-sounding voices in Indian languages. Bulbul V3 currently offers 35 voices in 11 languages and is optimized for low-bandwidth settings like call centres. Sarvam also offers smaller language models and APIs for speech-to-text and translation, targeting use cases like phone apps and local-language virtual assistants. The startup's goal is practical: to turn research into "compelling products" for everyday use in Indian enterprises and government. In sum, Sarvam AI is an Indian startup building AI for India – from OCR and voice bots to full-stack generative AI platforms.

Who Are The Founders of Sarvam AI

Sarvam AI was co-founded by Dr. Vivek Raghavan and Dr. Pratyush Kumar in 2023. Both founders have strong tech backgrounds focused on India:

• Dr. Vivek Raghavan – Co-founder. Raghavan was a key architect of India's digital public infrastructure (he worked at UIDAI, the Aadhaar authority, for over a decade). His experience includes large-scale government tech projects, giving Sarvam deep insight into India's needs.

• Dr. Pratyush Kumar – Co-founder and CEO. Kumar is a computer scientist (Ph.D. from ETH Zurich) and serial tech entrepreneur. He previously launched AI4Bharat (an open-source Indian language AI project) and PadhAI (an education app). He also has industry experience at Microsoft Research and IBM. As CEO of Sarvam AI, Kumar leads the company's "self-dependent AI" vision and frequently shares progress on social media.

Together, Raghavan and Kumar lead a growing team of researchers and engineers dedicated to India-focused AI. Sarvam's official About Us page highlights that "Vivek brings experience building India's digital public infrastructure, while Pratyush has led the country's open-source AI efforts across Indian languages". (The company is also actively hiring for new roles.) These founders are widely credited with steering Sarvam to its recent successes.

Funding & Reviews

Sarvam AI has attracted major venture capital backing. In December 2023, the startup announced a $41 million Series A round led by Lightspeed Venture Partners^[2], with participation from Peak XV Partners (formerly NEA) and Khosla Ventures. This strong funding round reflects confidence in Sarvam's vision, even though the company has kept its valuation private.

Press and users have also given Sarvam AI positive reviews. Technology media have highlighted Sarvam's India-first focus. For example, Sarvam "is making significant strides in India's AI landscape," noting Vision's lead on benchmarks and Bulbul's natural-sounding Indic voices. User feedback on social media has been enthusiastic. On X (formerly Twitter), one person exclaimed "Oh man wow" after using Sarvam's models. Notably, tech commentators like Deedy Das have publicly praised Sarvam. Das wrote that Sarvam "has the best text-to-speech, speech-to-text, and OCR models for Indic languages" and that their performance is "really valuable". Similarly, entrepreneur Pratik Desai applauded Bulbul: he said Sarvam's TTS has become his "go-to model" for Indic use cases, highlighting its affordability compared to ElevenLabs.

In short, Sarvam AI's performance on Indian-language benchmarks has generated widespread media attention and praise. Analysts and users agree that Sarvam's locally tuned models fill gaps left by global systems. These positive reviews, combined with strong investor backing, suggest Sarvam AI is a rising star in India's AI scene.

Benchmark Comparison: Sarvam AI vs Google Gemini vs ChatGPT

• Notes: A dash (—) indicates not reported.
• Sarvam Vision achieved 84.3% accuracy on the olmOCR-Bench (an Indian-language OCR test), surpassing Gemini 3 Pro and other models. On OmniDocBench v1.5 (a document analysis benchmark), Vision scored 93.3%. By contrast, Gemini 3 Pro scored lower than 84.3% on the OCR test, and GPT-4 (ChatGPT) was "significantly lower". In practical terms, Sarvam's model outperforms the larger global models on these India-focused tasks.

The table above illustrates why Sarvam AI's announcement made news: on these benchmarks it beats Google Gemini and ChatGPT on Indian documents. These tests involve reading mixed-language forms, tables and math, etc. Sarvam's training on Indian texts (22 languages) gives it an edge. (For voice, Sarvam's Bulbul V3 also showed lower error rates than global TTS systems on telephony audio, though we don't have a simple percentage to compare.) Overall, Sarvam AI leads on the specific tasks it targets.

Conclusion

Sarvam AI's recent achievements underscore India's growing AI capabilities. By focusing on local languages and real-world challenges, Sarvam's models outshine larger global systems on Indian tasks. The startup's success – from media acclaim to government partnerships (e.g., its AI collaboration with UIDAI for Aadhaar services) – highlights a shift toward "India-first" AI. For businesses and developers looking to leverage AI for Indian contexts, Sarvam AI provides both proof of concept and practical tools.

At GrowthJockey, we track these developments closely. If you're inspired by Sarvam AI's story and want to build or scale AI-powered solutions, GrowthJockey can help. We are a full-stack venture builder accelerating tech startups (including AI-driven firms) in India. Learn more on our GrowthJockey homepage about how we partner with entrepreneurs to bring world-class products to market.

FAQs

What does Sarvam AI do?

Sarvam AI builds AI solutions specially tuned for India. Its main products include an OCR/document AI (Sarvam Vision) and a text-to-speech engine (Bulbul V3), both trained on Indian language data. The company offers APIs for vision and speech tasks, aiming to help businesses and government digitize and automate local-language workflows. In essence, Sarvam AI provides "AI that works for India's linguistic mix".

Who is the CEO of Sarvam AI?

Dr. Pratyush Kumar is the co-founder and CEO of Sarvam AI. Kumar co-founded the startup with Dr. Vivek Raghavan and drives Sarvam's India-focused AI vision. (Vivek Raghavan is the other co-founder, with deep experience in India's public tech infrastructure.)

What is the salary of AI engineers at Sarvam AI?

Specific salaries at Sarvam AI are not publicly disclosed. However, industry reports suggest that mid-stage AI startups in India pay machine learning/AI engineers around ₹30–40 lakh per year on average. Sarvam AI's compensation is likely in line with these market rates (though exact figures would depend on role and experience).

Who invested in Sarvam AI?

Sarvam AI's investors include top venture funds. In its latest round (Dec 2023), Lightspeed Venture Partners led the funding, with support from Peak XV Partners (formerly NEA) and Khosla Ventures. This Series A (and earlier seed) raise totaled $41 million.

1. Sarvam AI's new OCR model Sarvam Vision "beats the likes of Gemini and ChatGPT in reading documents in Indian languages" - Link ↑
2. startup announced a $41 million Series A round led by Lightspeed Venture Partners - Link ↑