# Sarvam: India's Full-Stack Sovereign AI Platform Sarvam is building AI infrastructure for India. The company develops foundation models, APIs, and enterprise applications designed for Indian languages, accents, and cultural context. Everything is developed, deployed, and governed entirely in India. Founded in August 2023 by Dr. Vivek Raghavan (CEO) and Dr. Pratyush Kumar (CTO). Backed by $41M from Lightspeed, Peak XV Partners, and Khosla Ventures. ## What Sarvam Does Sarvam operates a full-stack sovereign AI platform spanning three layers: population-scale applications, state-of-the-art foundation models, and serving infrastructure. The platform supports 11 Indian languages and has handled 62M+ conversations with 99.9% uptime. ## Products ### Samvaad -- Conversational AI Agents Voice agents that sound human in every Indian language. Handle collections, customer support, sales, lead generation, KYC verification, and onboarding at scale. Agents connect to enterprise tools, use customer data, take action, and deliver insights across voice, WhatsApp, and web channels. - Sub-500ms latency - 100M+ conversations handled - 10x ROI for enterprise customers - Deploy in under 24 hours - One agent supports 11 languages simultaneously - Available across voice, WhatsApp, and web channels Use cases: lead generation (3x conversions), customer support (55% cost savings), KYC and verification (90% automated), collections, appointment booking, payment follow-ups, cart recovery. ### Studio -- Multilingual Content Transformation AI dubbing, translation, and subtitles in 11 Indian languages. Take one piece of content and reach every region -- video, documents, and audio. Ships 10x faster than manual workflows. Trusted by: Prime Minister's Office (Mann Ki Baat in 11 languages), IndiaAI/MeitY (national AI course), NPTEL (educational content localization), NCERT (curriculum video and textbook translation). ### Arya -- Enterprise AI Workflow Platform AI agents that automate complex enterprise workflows, from compliance reviews to customer operations, with full observability and zero vendor lock-in. Features include sovereign data residency, 360-degree visibility into agent decisions, full audit trails, and model-agnostic architecture. Use cases: compliance and risk review (insurance), customer operations, document processing, workflow automation. ### Akshar -- Document Digitisation A document digitisation platform that reads, understands, and extracts knowledge from real-world documents: scanned archives, handwritten notes, ancient scripts, complex tables, and dense Indic scripts. Handles complex layouts, Indic script conjuncts and matras, and preserves table structure. ## APIs and SDKs Sarvam provides developer APIs with clear documentation and a generous free tier. Available products: ### Text-to-Speech (Model: Bulbul v3) Natural, expressive voice synthesis in 11 Indian languages. Used for audiobooks, dubbing, assistive technology, and conversational AI. Pricing: Rs.30 per 10K characters. ### Speech-to-Text (Model: Saaras v3) Accurate transcription built for Indian accents, dialects, and real-world audio. Delivers accurate transcripts with diarization and handles background noise, multiple accents, and mid-sentence language switches. Pricing: Rs.30 per hour. ### Translation (Model: Mayura v1) Translates directly between Indian languages without using English as a bridge. Preserves meaning, tone, and cultural context across 10+ Indian languages. Pricing: Rs.20 per 10K characters. ### Document Intelligence / Vision (Model: Sarvam Vision) 3B parameter vision-language model for document digitisation, OCR, and visual understanding in Indian languages. Extracts text, tables, and structure from documents with precision. Pricing: Rs.1.50 per page. ### Large Language Models - Sarvam 105B: Flagship multilingual LLM for enterprise-grade Indian language applications. Free tier available. - Sarvam 30B: Fast, cost-efficient multilingual chat LLM with strong reasoning. Free tier available. ### Additional APIs - Transliteration: Convert text between Latin and native Indic scripts. Rs.20 per 10K characters. - Language Identification: Detect the language of any text snippet across Indian languages. Rs.3.50 per 10K characters. ## Foundation Models Sarvam develops its own foundation models trained on sovereign data: - Sarvam 105B: Flagship large language model for enterprise-grade Indian language applications. - Sarvam 30B: High-performance multilingual LLM optimized for Indian languages with strong reasoning. - Sarvam-M: Open-weight hybrid reasoning LLM for Indic languages (available on Hugging Face). - Bulbul v3: Natural, expressive text-to-speech across 11 Indian languages. - Saaras v3: Streaming speech recognition for 22 Indian languages with low-latency decoding and code-mixed support. - Sarvam Vision: 3B state-space vision-language model for document digitisation and OCR. - Mayura: Translation model handling colloquial language, code-mixing, and regional expressions. - Sarvam Translate: Open-weights translation model supporting 22 Indian languages. ## Customer Case Studies ### Tata Capital -- 3x increase in customer engagement Scaled multilingual voice agents across consumer loan products, breaking language barriers and deepening personalization. "Our partnership with Sarvam has enabled us to scale highly personalized, product and segment-specific conversations across the customer lifecycle. By embedding multilingual interactions across our consumer loan products, we are reaching more customers with greater relevance, breaking access barriers, and deepening engagement in a cost-effective manner." -- Shallu Kaushik, Chief Digital Officer, Tata Capital. ### SBI Life -- Millions of policy calls automated AI-powered voice agents handling policy inquiries, renewals, and claims in 10+ Indian languages. ### Ministry of Agriculture -- 50,000+ farmer feedback calls Conversational AI agent collecting structured citizen feedback for Farmer Field School training programs across Maharashtra. ## Key Metrics - 62M+ conversations handled - 11 Indian languages supported - 99.9% uptime for enterprise - 1.4B people the platform is built for - 22 official Indian languages covered by models - 100% of data stays in India - $41M raised in funding ## Why Sarvam - Sovereign by design: Build, deploy, and run AI with full control. Developed and operated entirely in India. DPDP compliant with data residency controls. - State-of-the-art models: Industry-leading models built for India's languages, culture, and context. - Human at the core: Forward-deployed engineers work alongside customer teams to deliver production-ready agents. - Secure and safe: AI that works on-prem, in the cloud, or at the edge. Available wherever needed. ## Languages Supported Sarvam models and products support major Indian languages including Hindi, Tamil, Telugu, Bengali, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, and more. Speech recognition (Saaras v3) covers 22 Indian languages. Translation models support direct translation between Indian languages without requiring English as an intermediary. ## Company - Tagline: "AI for all from India" - Category: Full-stack sovereign AI - Headquarters: India - Founded: August 2023 - Founders: Dr. Vivek Raghavan (CEO), Dr. Pratyush Kumar (CTO) - Investors: Lightspeed, Peak XV Partners, Khosla Ventures - Website: https://www.sarvam.ai - Developer Dashboard: https://dashboard.sarvam.ai - Documentation: https://docs.sarvam.ai