MeitY issues AI agent tender for DigiLocker

India's Ministry of Electronics and Information Technology (MeitY) issued a tender on August 7, 2026 to recruit AI/ML professionals for a Unified AI Team that will design, develop and deploy AI services across Digital India platforms such as UMANG and DigiLocker. The tender, released by the National e-Governance Division (NeGD), invites AI-capable agencies to provide staff for seven key domains, including inference server setup, cognitive analytics, computer vision, conversational AI, generative AI, data engineering, and responsible AI governance.

The tender document can be downloaded here: fixtures. It outlines the AI services that the selected agencies are expected to handle, as well as the technical infrastructure they must implement.

Scope of the Tender

NeGD intends to establish an in-house server to deploy and serve open-source models behind stable APIs for use by any NeGD or DIC project. The infrastructure will run on approved government cloud services such as NIC or MeghRaj. The RFE states that the system architecture should support switching between two model categories:

• Commercial models from OpenAI, Anthropic, Google, AWS and Azure.
• Open-source LLMs such as LLaMA and Mistral to ensure security, compliance and data localisation.

The RFE notes that the listed activities are illustrative and additional tasks may include vulnerability remediation or evidence generation for compliance.

The tender covers seven domains:

• Setting Up Inference Server
• Cognitive Analytics & Decision Intelligence
• Computer Vision & Document Intelligence
• Conversational & Assistive AI
• Generative AI & LLM Engineering
• Data Engineering & MLOps
• Responsible AI & Governance

Technical Requirements

The RFE lists six AI services that the deployed AI agencies are supposed to work on:

• Document Summarisation and Classification
• Multilingual Conversational AI Chatbots
• Voice-based Form Filling and Web Navigation
• OCR Processing of Documents
• Live Face Detection for ID Verification
• AI Agents for Government Applications

The AI agents will be built for government applications such as DigiLocker, using tools such as LangGraph or CrewAI. These agents are intended to securely invoke approved APIs and tools to complete transactions while maintaining audit trails and human checkpoints for irreversible actions.

The RFE outlines the following agentic actions:

• Building agent-based AI capabilities to automate and assist government workflows, starting with DigiLocker.
• Developing modular, reusable agent frameworks for adoption across ministries and platforms.
• Designing agents to be discoverable and integrable by third-party developers.
• Incorporating human-in-the-loop validation for sensitive actions.
• Maintaining comprehensive audit trails, access logs and traceability for all agent decisions and API calls.
• Demonstrating end-to-end reference implementations, such as agent-driven document retrieval and verification via DigiLocker APIs.
• Developing specialised agent capabilities for querying large databases or structured data tables, enabling natural-language query processing and analytical insights.

According to the RFE, these agents are to function as secure, policy-aligned digital assistants capable of understanding user intent, orchestrating tasks and acting responsibly.

The tender also highlights that the AI services will support multilingual interactions, voice-driven form filling, and live face detection for identity verification, thereby strengthening KYC processes and reducing fraud.

The RFE documents the expected deliverables and technical specifications, providing a clear roadmap for agencies interested in participating in this national AI initiative.