NeGD is currently inviting applications for AI Engineer positions on a contractual basis, initially for three years, with possible extensions based on project requirements.

Responsibilities

• Develop and maintain scalable backend and frontend components for AI-based web applications using Python, Node.js and React.js, following clean, maintainable, production-ready coding practices.
• Design, build and secure RESTful and GraphQL APIs; integrate AI services with existing platform API layers following documented conventions, without bypassing the API layer or modifying underlying database schemas.
• Build document intelligence pipelines covering OCR, layout analysis, document-type classification from content, entity and key-value extraction, and table extraction from multi-page, scanned, multilingual and partially legible documents; cross-reference extracted values against structured form data and surface material discrepancies as non-blocking advisories.
• Build retrieval-augmented generation pipelines over policy, scheme, regulatory, legal and precedent corpora, including clause-aware chunking, embedding model selection, vector store design, hybrid sparse-dense retrieval, reranking and query rewriting, with retrieval quality measured rather than assumed.
• Implement source attribution end to end, so that every generated statement carries the provision, document section or data record it derives from and generated content remains visually distinguishable from primary data.
• Build cross-document synthesis and summarisation that compresses long reports and multi-round submission histories into structured, source-linked summaries, and detects semantic inconsistency within and across documents.
• Build conversational assistants with intent classification and dialogue state management, supporting both public unauthenticated access and authenticated role-aware access, with responses restricted to the data and scope permitted for each user role.
•  Implement multilingual and voice interaction across Indian languages, including speech recognition and synthesis, transliteration and script handling, and integration with national language infrastructure.
•  Implement confidence-based abstention and human handoff, so that low-confidence queries escalate to a helpdesk or officer with the query, conversation history and topic context pre populated.
• Build agentic workflows with policy-bounded multi-step planning, tool and function calling, state management and explicit human-in-the-loop checkpoints, such that no agent action progresses a workflow, updates a record or substitutes for official judgement.
•  Build assisted drafting capabilities that generate strictly from approved precedent and supplied reference material, and flag missing input rather than producing unsupported content.
• Evaluate and select foundation models capability by capability, treating open-weight and sovereign models as the default position, assessing on retrieval grounding, citation fidelity, hallucination rate, latency and cost rather than general benchmark performance, and documenting justification for any departure.
• Deploy and operate model inference services including quantisation, batching, caching, GPU sizing and latency management against agreed performance targets.
• Build evaluation harnesses for generative components measuring faithfulness, answer relevance and context precision and recall, with regression suites over a curated query set and red-teaming for prompt injection and unsafe output.
• Implement guardrails including grounding constraints, refusal and abstention behaviour, prompt injection and data-exfiltration defence, personal-data redaction, and role-based filtering of AI output.
• Work with relational, NoSQL and vector databases, ensuring optimised queries, scalable data access and reliable performance.
• Implement application security best practices including authentication, authorisation, role-based access control, secrets management and secure handling of sensitive data; support security testing by CERT-In empanelled auditors and remediate findings.
• Write every AI output, its confidence characterisation and the corresponding human decision into the platform audit trail, and ensure each module remains independently deployable and disableable without impacting the core platform or other modules.
• Develop and manage CI/CD pipelines and containerised deployments; implement observability, alerting and cost governance for model and API consumption.
• Apply Test-Driven Development principles with adequate test coverage; manage the codebase using Git with branching and review workflows; participate actively in code reviews.
• Follow Agile/Scrum practices and contribute to sprint planning, stand-ups, demonstrations and technical documentation including API specifications and runbooks.
• Mentor junior developers, share knowledge within the team, and contribute to other NeGD projects as assigned.

Important Links

About National e-Governance Division (NeGD)

The National e-Governance Division (NeGD) is an independent business division under the Digital India Corporation, Ministry of Electronics and Information Technology. NeGD has been playing a pivotal role in supporting MeitY in Programme Management and implementation of e-Governance projects and initiatives undertaken by various Ministries/ Departments, both at the Central and State levels.

NeGD has been spearheading several innovative initiatives under the aegis of the Digital India Programme. Those have been developed keeping the vision areas of Digital India at the core- providing digital infrastructure as a core utility to every citizen, governance and services on demand and in particular, digital empowerment of the citizens of our country; some of these initiatives include DigiLocker, UMANG, Poshan Tracker, OpenForge Platform, API Setu, National Academic Depository, Academic Bank of Credits, Learning Management System.