Role Purpose

The Assistant Manager / Manager AI Security is responsible for leading the design, assessment, and validation of secure AI systems across their lifecycle. The role combines AI security architecture, secure-by-design advisory, AI security tooling, and AI red teaming to help organisations deploy AI safely, securely, and at scale.

Key Responsibilities

1 AI Security Architecture & Secure-by-Design Advisory

Lead and review AI security architecture for:

- AI platforms, pipelines, and MLOps environments
- GenAI applications (LLMs, copilots, agents, RAG pipelines)

Define and embed securebydesign principles across:

- Data ingestion and training pipelines
- Model hosting, inference APIs, and AI agents
- Identity, access control, and isolation for AI systems

Advise clients on defenseindepth approaches aligned to:

- Zero Trust (NHI, IAM) for AI
- Enterprise cloud and application security patterns
- AIspecific threat models

2 AI Security Across the AI Lifecycle

Data & Training Phase

- Assess risks related to:
- Data poisoning, bias manipulation, and training data leakage
- Weak provenance, lineage, and thirdparty data dependencies

- Define controls for secure data handling and integrity validation.

Model Development & FineTuning

- Perform AI threat modeling covering:
- Model inversion and extraction
- Backdoored or trojaned models

- Review secure MLOps pipelines, CI/CD controls, and isolation mechanisms.

Deployment & Inference

- Secure AI runtime environments across cloud, containerized, and hybrid setups.
- Define controls for:
- Prompt injection resistance
- Output filtering, guardrails, and misuse prevention
- Abuse scenarios in GenAI and agentbased systems

Monitoring & Operations





- Support AI Security Posture Management (AISPM) and continuous assurance.

AI Red Teaming & Adversarial Testing

Plan and execute AI Red Team engagements, including:

- Prompt injection and jailbreak testing
- Model misuse, data exfiltration, and policy bypass scenarios
- AIenabled attack simulations

Apply industry frameworks such as:

- MITRE ATLAS
- OWASP Top 10 for LLM Applications
- OWASP Agentic AI
- NIST AI RMF

Design custom adversarial scenarios aligned to client context and risk appetite.

Translate red team findings into actionable remediation roadmaps for engineering and leadership teams.

Engagement Leadership & Practice Contribution

- Manage small to midsized AI security engagements or workstreams.
- Act as a subjectmatter resource for AI security across pursuits and proposals.
- Mentor junior team members on AI security concepts and delivery.
- Contribute to thought leadership, playbooks, and internal enablement on AI security and red teaming.

Required Skills & Experience

Solid understanding of cybersecurity fundamentals (IAM, AppSec, Cloud Security, Threat Modeling).

Practical understanding of AI/ML and GenAI architectures, including:

- Model training vs inference
- LLMs, agents, and RAG patterns

Experience with:

- AI security assessments or testing
- Secure MLOps / DevSecOps pipelines
- Offensive security or adversarial testing (preferred)

Preferred Certifications

- CAI/ML Pen Certificate (Certified AI/ML Penetration Testing)
• CEH, OSCP, or equivalent offensive security certification
• GIAC certifications (GPEN, GWAPT, GCTI)
• Cloud Security certifications (AWS / Azure / GCP)
• AI Governance / Responsible AI certifications or NIST AI RMF training
• CISSP / CCSP (advantageous)

📌 VAPT + AI Security ( Pentesting with AI Experience) (Pune)
🏢 KPMG Assurance and Consulting Services
📍 Pune