Overview

NeGD is building a portfolio of AI-driven digital products for citizens at national scale conversational assistants, agentic platforms, document and decision-support systems, identity and verification services, fraud-detection solutions, and other AI-enabled government platforms.

We are looking for a Senior Security Engineer to secure these systems end to end: someone who is equally comfortable identifying vulnerabilities in a conventional web application or API, reviewing authentication and cloud configurations, and attacking an LLM, RAG pipeline, ML model, or autonomous agent.

This is a hands-on engineering role. The ideal candidate does not only define security policies, coordinate audits, or review reports. They can personally inspect code and architecture, build threat models, use offensive and defensive security tooling, reproduce vulnerabilities, write scripts and test harnesses, configure security controls, and work directly with engineers to remediate findings.

The role requires depth in established cybersecurity disciplines — application security, API security, identity, cloud and infrastructure security, secure SDLC, penetration testing, cryptography, vulnerability management, and incident response — together with practical understanding of the emerging attack surface introduced by AI systems.

The selected engineer will work alongside software engineers, AI/ML engineers, MLOps and DevOps/ SRE teams, product managers, AI safety specialists, and government stakeholders to build systems that are secure by design and resilient against both traditional and AI-specific attacks.

Roles and Responsibilities

- Conduct hands-on security assessment of web applications, APIs, microservices, mobile/backend services, cloud deployments, containers, and supporting infrastructure, using a combination of manual testing and automated security tooling.
- Perform application and API penetration testing covering authentication and authorisation failures, injection vulnerabilities, insecure business logic, SSRF, deserialisation issues, file-handling vulnerabilities, privilege escalation, session weaknesses, secrets exposure, and other OWASP-class vulnerabilities.
- Conduct security reviews of AI-enabled applications, including LLM, RAG, agentic AI, predictive ML, computer vision, identity-verification, and fraud-detection systems, identifying risks introduced by models, prompts, retrieval layers, tools, data pipelines, and integrations.
- Design and execute LLM and agentic-AI red-team exercises covering direct and indirect prompt injection, jailbreaks, system-prompt leakage, sensitive-data disclosure, insecure tool invocation,excessive agency, permission bypass, output manipulation, cross-user data leakage, and policy-boundary violations.
- Assess RAG security including poisoning or manipulation of retrieved content, malicious document ingestion, retrieval-based prompt injection, source-integrity weaknesses, access-control failures in vector stores, sensitive-data retrieval, and trust-boundary failures between retrieval and generation components.
- Test traditional and machine-learning systems for relevant adversarial ML threats, including adversarial inputs, model extraction, model inversion, membership inference, data poisoning, training-data leakage, and abuse of prediction or inference APIs.
- Build reusable scripts, attack harnesses, payload libraries, test datasets, and automated security checks in Python, Bash, or equivalent languages so security testing can be reproduced and integrated into engineering workflows.
- Perform threat modelling for conventional and AI-enabled systems using frameworks such as STRIDE, attack trees, MITRE ATT&CK;, MITRE ATLAS, OWASP ASVS, OWASP API Security Top 10, and OWASP guidance for LLM and generative-AI applications.
- Review architecture and code for secure implementation of authentication, authorisation, session management, API security, input validation, output handling, file processing, secrets management, logging, encryption, and data isolation.
- Review OAuth 2.0, OpenID Connect, SAML, RBAC, ABAC, service-to-service authentication, API keys, tokens, and workload identities, and identify privilege-escalation or access-control weaknesses across applications and AI services.
- Integrate and operate security testing within the software delivery lifecycle, including SAST, DAST, SCA, secrets scanning, container and image scanning, IaC scanning, dependency vulnerability management, and security gates within CI/CD pipelines.
- Conduct manual code review and configuration review where automated tooling is insufficient, particularly for high-risk services, custom authentication flows, sensitive-data processing,



and AI orchestration logic.
- Assess cloud, Kubernetes, container, network, API-gateway, WAF, secrets-management, logging, and model-serving configurations for exploitable misconfigurations and insecure defaults, working closely with DevOps/SRE and MLOps engineers.
- Review the security of AI model and data supply chains — training and evaluation data, model artefacts, open-source models, third-party APIs, packages, containers, model registries, prompt templates, embeddings, plugins, tools, and external knowledge sources.
- Define and help engineers implement practical mitigations for AI attacks, including prompt and context isolation, privilege minimisation, tool allow-listing, deterministic security controls outside the model, input/output validation, secure retrieval, sandboxing, human approval for high-impact actions, and monitoring for abuse.
- Verify that security controls work in practice by attempting to bypass them rather than relying solely on design documentation or vendor claims.
- Reproduce findings reported by VAPT providers, automated scanners, researchers, or internal teams; determine exploitability and severity; and work directly with engineering teams until remediation is technically verified.
- Support security incident investigation and root-cause analysis, including examination of application logs, cloud and infrastructure telemetry, authentication events, AI interaction logs, model/tool activity, and indicators of compromise or abuse.
- Contribute secure coding patterns, threat models, red-team playbooks, testing utilities, security checklists, and reusable engineering controls that can be adopted across multiple government AI projects.
- Stay current with developments in application security, cloud security, offensive security, adversarial machine learning, LLM security, agent security, and AI supply-chain security, and translate relevant research into practical engineering controls and tests.

Required Profile

Educational Qualifications

- Bachelor's degree in Computer Science, Information Technology, Cybersecurity, Electronics, or a related engineering discipline.
- A Master's degree in Information Security, Cybersecurity, Computer Science, Artificial Intelligence, or a related field is an added advantage.
- Strong demonstrable practical security capability may be considered alongside non-traditional educational backgrounds.

Experience and Skills — Essential

- Minimum 5 years of hands-on cybersecurity experience, with substantial experience in application security, penetration testing, product security, offensive security, cloud security, or security engineering. Depth and quality of practical work matter more than years alone.
- Demonstrated ability to personally conduct security testing rather than only manage security vendors, compliance programmes, or audit processes.
- Strong practical knowledge of web and API security, including the OWASP Top 10, OWASP API Security Top 10, authentication and authorisation vulnerabilities, session management, access-control flaws, injection classes, SSRF, secrets exposure, and business-logic abuse.
- Hands-on proficiency with security tooling such as Burp Suite, OWASP ZAP, Nmap, Semgrep, SonarQube, Snyk, Trivy, Gitleaks, Dependency-Track, Metasploit, or equivalent tools, with the ability to go beyond scanner output and manually validate findings.
- Ability to read and reason about application code in at least one commonly used language such as Python, JavaScript/TypeScript, Java, Go, or C#, and sufficient coding ability to create security scripts, proof-of-concepts, test harnesses, or automation.
- Practical understanding of cloud and container security, including Docker, Kubernetes, IAM, secrets management, network segmentation, TLS, cloud storage, API gateways, WAFs, workload identity, and common cloud misconfiguration patterns.
- Solid understanding of identity and access security — OAuth 2.0, OpenID Connect, SAML, RBAC/ ABAC, token security, service identities, API authentication, privilege boundaries, and least-privilege design.
- Experience with secure SDLC and DevSecOps practices, including threat modelling, secure design review, code review, SAST, DAST, SCA, secrets scanning, vulnerability management, and integration of security controls into CI/CD.




- Working understanding of cryptographic security fundamentals including TLS, encryption at rest, key and certificate management, hashing, signing, PKI, and appropriate use of secrets-management systems.
- Demonstrated understanding of the architecture of modern AI applications — LLM APIs, system and user prompts, RAG, embeddings and vector databases, agents and tool calling, model serving, ML pipelines, and third-party/open-source models.
- Hands-on experience testing at least some AI-specific security risks, such as prompt injection, jailbreaks, sensitive-information disclosure, insecure tool use, RAG manipulation, adversarial inputs, model extraction, or data poisoning.
- Familiarity with frameworks and guidance such as MITRE ATT&CK;, MITRE ATLAS, OWASP ASVS, OWASP Testing Guide, and OWASP security guidance for LLM and generative-AI applications.
- Ability to distinguish between weaknesses that should be addressed through model behaviour or guardrails and security boundaries that must be enforced deterministically in application code, identity systems, infrastructure, or architecture.
- Strong analytical and communication skills, with the ability to explain an exploit to an engineer, a technical risk to an architect, and the business impact of a vulnerability to a non-security stakeholder.
- Outcome-driven and comfortable working as an independent hands-on specialist while collaborating closely with software, AI/ML, MLOps, DevOps/SRE, product, and governance teams.

Experience and Skills — Desirable

- Experience conducting structured red-team exercises against production or production-like LLM, RAG, multimodal, or agentic AI systems.
- Experience with adversarial machine learning frameworks or techniques, including ART, Foolbox, CleverHans, or equivalent, and attacks such as model extraction, inversion, membership inference, poisoning, or adversarial examples.
- Familiarity with AI-security testing and evaluation tools, emerging LLM red-team frameworks, guardrail testing frameworks, or experience creating custom attack harnesses where available tools are insufficient.
- Experience securing vector databases, ML pipelines, model registries, GPU/model-serving infrastructure, AI APIs, and model or data supply chains.
- Experience with AWS, Azure, GCP, sovereign cloud, on-premise Kubernetes, or comparable enterprise infrastructure.
- Experience with security monitoring, detection engineering, incident response, SIEM, audit logs, and investigation of suspicious activity across application and cloud environments.
- Prior work in government, BFSI, critical infrastructure, identity platforms, citizen-scale digital services, or other regulated/high-assurance environments is a plus.
- Familiarity with Government of India security and data-protection requirements, including MeitY security guidance, CERT-In requirements, STQC processes, DPDPA 2023, and relevant Government of India data-classification requirements.
- Relevant certifications such as OSCP, OSWE, GWAPT, CISSP, CSSLP, CKS, cloud-security certifications, or equivalent are desirable, but practical security capability will be given greater weight than certifications alone.
- Security research, responsible disclosure, CVEs, bug-bounty findings, CTF performance, open-source security contributions, conference presentations, technical blog posts, attack tools, or demonstrable personal security projects are strongly valued.

Selection Process Shortlisted candidates will be expected to demonstrate practical security depth during the interview rather than relying solely on certifications or theoretical knowledge.

Candidates should be prepared to discuss two or three security assessments, vulnerabilities, red-team exercises, or security-engineering projects that they personally worked on, including the system architecture, attack path, tools and techniques used, exploitability assessment, remediation, and how the fix was validated.

The assessment may include a practical or scenario-based exercise covering both:

Conventional security, such as analysing an application/API architecture, reviewing code or configuration, identifying an authentication or authorisation flaw, interpreting scanner findings, or developing a penetration-testing approach; and

AI security, such as threat-modelling an LLM/RAG or agentic system, designing prompt-injection or data-exfiltration tests, analysing unsafe tool use, identifying trust boundaries, or proposing and validating controls against an AI-specific attack.

The selection process will place particular weight on whether the candidate can personally investigate, exploit, automate, remediate, and verify security issues across both traditional software systems and modern AI applications.

📌 Senior Security Engineer (New Delhi)
🏢 National e Governance Division
📍 New Delhi