Build AI · Builders

Deploy India's sovereign language model locally or on cloud. Official guidance from IndiaAI including API access and fine-tuning instructions.

Krutrim's API for building Indian-language applications. Supports 22 languages with low-latency inference on Indian cloud infra.

The Government of India's IndiaAI Mission provides subsidised GPU compute to Indian AI startups, researchers, and institutions through empanelled cloud providers. Apply via the IndiaAI portal. Allocated 10,000+ GPUs specifically for Indian organisations — much cheaper than AWS/GCP for qualifying entities.

Krutrim (by Ola Founder Bhavish Aggarwal) is India's first LLM trained on Indian languages and context. Free API access for developers. Supports Hindi, Tamil, Telugu, Kannada, Bengali, Marathi, and more. Lower latency for Indian users vs US-based models. India-based data residency.

IIT Madras AI4Bharat's open-source translation model supporting all 22 scheduled Indian languages. Outperforms Google Translate on Indic language pairs. Free API and model weights on HuggingFace. Used by government applications, edtech, and media companies for Indian-language content.

MeitY's Bhashini platform provides free NLP APIs: ASR (speech-to-text), TTS (text-to-speech), translation, and OCR for all Indian languages. Free for Indian developers and organisations. Powers government apps, digital inclusion initiatives, and private sector integrations. Sign up via bhashini.gov.in.

Meta's Llama 3.2 is now free to run on Google Colab T4 GPU — no cost. Step-by-step guide to running a local LLM in India: model download, quantisation with llama.cpp for smaller VRAM, and building a simple chat interface. Works on Indian internet speeds with llama.cpp's GGUF format.

Google Cloud's Mumbai (asia-south1) and Delhi (asia-south2) regions support Vertex AI, Gemini API, Vision AI, and Speech-to-Text with Indian data residency. ₹20,000 free credits for new accounts. Relevant for DPDP Act compliance — keeps data within India for regulated industries.

Azure OpenAI Service is available in the India Central region (Pune). Deploy GPT-4o, GPT-3.5-turbo, and text-embedding models with Indian data residency. Free 12-month credits for students via Azure for Students. Commonly used by Indian banks, insurance companies, and government for DPDP compliance.

Hugging Face Spaces gives every user 2 free CPU instances to deploy Gradio/Streamlit AI apps. No server setup, no cost. Used by thousands of Indian AI researchers to demo models. Upgrade to GPU Spaces for ₹1,500/month. The fastest way to share an AI prototype without DevOps.

Vercel's free tier supports AI-powered Next.js apps. The AI SDK connects to OpenAI, Anthropic, Google Gemini, and Mistral with streaming support. Free hosting on Vercel's global CDN with Mumbai edge location for low latency in India. This site is built with it.

Ollama lets you run Llama 3, Mistral, Phi-3, and Gemma locally with one command. Works on an 8GB RAM laptop. No internet required after download. Perfect for Indian developers in areas with unreliable connectivity, or for privacy-sensitive applications where data can't leave the device.

Sarvam AI (Indian startup) provides speech and language AI APIs for Indian languages — ASR, TTS, translation, and transliteration. Built by ex-Google and ex-Anthropic researchers. DPIIT-recognised startup. Competitive pricing in INR. Better than global models on Hindi, Tamil, Telugu vernacular.

E2E Networks is an Indian-listed GPU cloud provider with NVIDIA A100/H100 GPUs at 40-60% lower cost than AWS. Data stored in India. Accepts Indian payment methods (UPI, NetBanking). Used by Krutrim, Indian AI startups, and academic labs for model training and fine-tuning.

Google Colab gives you a free NVIDIA T4 GPU in the browser. No setup, no hardware. Perfect for running ML models, Jupyter notebooks, and fine-tuning experiments from any Indian city. Works on 4G.

Hugging Face Spaces lets you host any ML model as a web app for free. 100,000+ open models including Llama, Mistral, and Whisper. India's AI builders use it for rapid prototyping without cloud bills.

Ollama lets you run Llama 3, Mistral, Gemma, and Phi-3 locally on Mac or Windows. No internet needed after download. Ideal for Indian developers with data privacy concerns or unreliable connections.

Groq's free API tier runs Llama 3, Mixtral, and Gemma at 500+ tokens/second — faster than any other free service. Indian builders use it as a fast, cost-free alternative to OpenAI for prototypes.

Google's Gemini 1.5 Flash is free up to 1,500 requests/day. Supports 1M token context, image understanding, and code generation. The best free LLM API for Indian developers building production apps.

OpenAI's free Playground lets you experiment with GPT-4o, prompt engineering, and function calling. $5 free credits on signup. Essential for understanding how LLM APIs work before building.

Krutrim Cloud provides Indian-sovereign GPU inference for Llama, Mistral, and Krutrim's own models. Data stays in India. Competitive pricing vs AWS/Azure. Built by Ola's Bhavish Aggarwal.

Bhashini is India's national language AI platform — free APIs for speech-to-text, text-to-speech, and translation across 22 official Indian languages. Built by MeitY, free for Indian developers.

IIT Madras's AI4Bharat provides free open-source models: IndicBERT (text), IndicTrans2 (translation), Dhruva (speech), and Sarvam (voice). All trained on Indian language data. API access is free.

LangChain is the standard framework for building LLM-powered applications: chatbots, RAG systems, agents, and pipelines. Free, open-source, extensive docs. Start with the Python quickstart.

Vercel's free AI SDK connects Next.js or React apps to any LLM (OpenAI, Gemini, Anthropic, Groq) in minutes. Handles streaming, tool calls, and multi-step agents. Indian developers use it with Krutrim too.

E2E Networks is India's GPU cloud provider — A100, H100, and RTX 4090s at 40–60% lower cost than AWS/Azure. Ideal for Indian startups and researchers who need training compute inside India.

Replicate hosts 10,000+ open-source models as APIs. Run Stable Diffusion, Llama, Whisper, and more with a single API call. Pay-per-use, no server management. Good for Indian developers testing ideas.

Kaggle gives 30 hours/week of free GPU (P100/T4) with 100GB storage. India has one of Kaggle's largest user bases. Use it for competitions, learning, and running ML experiments.

Claude 3.5 Haiku is free on the Anthropic API (limited calls/day). Excellent for document analysis, long-context tasks, and coding. India-based developers use it as a GPT-4o alternative.