Job Summary & Responsibilities :

Own the security of our AI and GenAI systems end to end from training data to production LLM apps and agents.

Our models handle patient, clinical trial, and research data, so every control must meet GxP, HIPAA, GDPR and the EU AI Act. You'll work with data science, engineering, legal and compliance to keep AI safe, auditable, and fast to ship.

Key responsibilities :

- Design and run security architecture reviews and threat models for AI/ML, LLM and agentic applications (STRIDE, MITRE ATLAS, OWASP Top 10 for LLMs).

- Build guardrails against prompt injection, jailbreaks, data leakage, insecure output handling and excessive agent permissions.

- Lead AI red-teaming and adversarial testing of models, RAG pipelines and agents; track findings to closure.

- Secure the ML lifecycle : training data provenance, model registry integrity, model signing, supply-chain scanning of open-source models and packages.

- Define controls for LLM gateways and API access : authentication, rate limiting, logging, PII/PHI redaction and content filtering.

- Embed AI security checks into CI/CD and MLOps pipelines (shift-left scanning, policy-as-code).

- Monitor production AI systems for abuse, drift, and anomalous behavior; integrate with SOC/SIEM and incident response.

- Translate regulations and frameworks (NIST AI RMF, ISO/IEC 42001, EU AI Act, GxP) into technical controls and evidence for audits.

- Review third-party AI vendors and SaaS tools for security and data-handling risk.

- Write standards, reference architectures and secure-by-default patterns, coach engineering and data science teams.

Required qualifications :

- Bachelor's or Master's in Computer Science, Cybersecurity or a related field.

- 8 + years in application, cloud, or product security engineering.

- 2 + years hands-on securing ML/AI systems, including at least one production GenAI or LLM application.

- Strong grasp of LLM attack classes : prompt injection, data poisoning, model extraction, membership inference, insecure plugins/tools.

- Working knowledge of OWASP Top 10 for LLM Applications, MITRE ATLAS and NIST AI RMF.

- Cloud security depth in at least one of AWS, Azure or GCP (IAM, network isolation, KMS, private endpoints).

- Coding ability in Python plus one of Go, Java, or TypeScript; able to build security tooling, not only review it.

- Experience with threat modeling, secure SDLC and CI/CD security.

- Clear written and verbal communication with engineers, data scientists and compliance stakeholders.

Preferred qualifications :

- Pharma, life sciences or healthcare background; familiarity with GxP, 21 CFR Part 11, HIPAA and GDPR.

- Hands-on AI red-teaming or published research, CVEs, or conference talks on AI security.

- Experience with agentic systems and tool-calling security (MCP servers, function calling, least-privilege agents).

- Exposure to ISO/IEC 42001 or EU AI Act readiness work.

- Certifications : CISSP, CCSP, OSCP, GIAC (GPEN, GCSA, GMLE), or cloud security specialty certs (AWS Security Specialty, Azure SC-100).

Tech stack and tools :

- Cloud and ML platforms - AWS (SageMaker, Bedrock), Azure (Azure OpenAI, Azure ML), GCP (Vertex AI), Databricks

- LLM frameworks - LangChain, LlamaIndex, OpenAI / Anthropic APIs, vector databases (Pinecone, pgvector, OpenSearch)

- AI security and red teaming - Garak, PyRIT, Promptfoo, Protect AI, Lakera, NeMo Guardrails, Azure AI Content Safety

- AppSec and supply chain - Snyk, Semgrep, Checkmarx, SBOM tooling, model scanning (ModelScan), Sigstore

- Infrastructure - Kubernetes, Terraform, OPA/policy-as-code, GitHub Actions or Azure DevOps

- Detection and response - Splunk or Microsoft Sentinel, CSPM/CNAPP (Wiz, Prisma Cloud)