Senior Product Security Engineer, AI ML

Leads multi-product security efforts, develops the ProdSec team and mentors engineers, coordinating incident response, and shaping DevSecOps practices, with primary focus on securing Norstella’s agentic SDLC platform (Forge) and its supporting knowledge-base ecosystem (Sage). This engineer owns the security posture of the autonomous AI agent chain end-to-end: the platform itself, the prompts and orchestration layer, the proprietary domain data the agents consume, and - critically - the source code the agents produce. Acts as the senior security partner to the Forge engineering team, embedding security guardrails into every phase of the autonomous SDLC (Intake * Architect * Threat-Modeling * Dev-Planning * Dev * QA) so that AI-generated code meets the same compliance, audit, and resilience standards as human-written code in HIPAA, GDPR, SOC 2, and FDA 21 CFR Part 11 environments., *Leads threat modeling across complex architectures, including agent-to-agent communication graphs, tool-use surfaces, MCP server boundaries, and prompt/RAG pipelines, and mentors engineers in risk analysis practices.

*Leads secure development training across product groups - including AI-assisted and AI-generated coding workflows - and evolves standards based on industry trends, OWASP LLM Top 10, and emerging agentic-AI threat research.

*Leads vulnerability management for multiple products, setting priorities and ensuring timely resolution, and tunes the Forge Security Agent’s SAST/SCA gating, severity thresholds, and triage-LLM classification logic to keep auto-remediation loops both effective and safe.

*Leads supply chain risk assessments and drives API governance practices across product portfolios, with explicit ownership of the AI/ML supply chain - foundation model providers (Anthropic, others), Auth0, Snyk, Azure DevOps NPM feeds, vector databases, and the Sage KB seed data - including concentration-risk monitoring and contingency planning.

*Leads security integration across CI/CD and multi-cloud platforms, mentoring engineers on automation, and ensures every agentic run produces a reconstructible audit trail sufficient to demonstrate reasonable judgment to SOC 2, HIPAA, GDPR, and Life Sciences auditors.

*Technical IR lead overseeing cross-product incidents, coordinates disclosure processes, and advises leadership, including incidents involving prompt injection, model jailbreak, agent privilege escalation, sensitive data leakage from KB stores, and compromised AI-generated code reaching production.

*Leads AI/ML model security practices across multiple teams and establishes safe design principles, including least-privilege tool-use, secrets handling for agent runtimes, PII/PHI sanitization in prompts and outputs, model-output validation, and red-team / adversarial-input programs targeting the Forge agent chain.

*Leads SSDLC governance initiatives, mentors teams on embedding controls, and validates adherence across product lines, with explicit accountability for the agentic SSDLC: ensuring the Threat Model Production Agent’s output is consumed downstream, that the Forge Security Agent’s fail-open posture is risk-tiered appropriately, and that human-in-the-loop checkpoints are placed where judgment is genuinely required.

*Designs and operates continuous security evaluation pipelines that benchmark agent runs across safety, compliance, and code-quality dimensions, surfacing regressions in real time.

*Defines acceptance criteria and security guardrails for AI-generated code and agent outputs, ensuring policy compliance and alignment with business intent before production promotion.

*Partners with Agentic Engineering to instrument observability, logging, and tracing across the full agent execution graph so every decision an agent makes is auditable, explainable, and defensible to regulators.

*Holds a BA/BS degree or equivalent experience; certifications such as CISSP, OSWE, or cloud security specialties preferred. AI/ML security credentials (e.g., AI Security Institute, MLSecOps, or equivalent) a strong plus.

*Typically 4+ years of experience or equivalent leadership across multi-product security programs, with at least 1-2 of those years applied to AI/ML or LLM-powered systems in production.

*Strong knowledge of managing SDLC risks across multiple teams and mentoring others in secure design, including agentic and AI-assisted SDLC patterns where code authorship is partially delegated to LLM agents.

*Proficient in enterprise vulnerability management, guiding priorities and automation strategies, including LLM-mediated triage and auto-remediation loops.

*Strong knowledge of integrating security into complex CI/CD and cloud environments, including containerized agent runtimes, GPU-accelerated inference endpoints, and serverless agent orchestration on AWS, GCP, or Azure.

*Strong knowledge of secure development practices and setting engineering-wide standards.

*Proficient in designing and reviewing encryption and data protection strategies.

*Proficient in automation in product pipelines and mentoring teams on efficiency.

*Proficient in AI/ML security governance across multiple teams, including OWASP LLM Top 10, prompt injection / jailbreak defense, RAG poisoning, model-output sanitization, and adversarial testing of agent chains.

*Proficient in SSDLC governance, leading implementation and mentoring peers on embedding security into development processes, with demonstrated ability to apply SSDLC controls to non-deterministic, agent-driven development workflows.

*Hands-on experience with LLM-powered applications (Claude, GPT-4, Gemini, Llama, or equivalent), including prompt engineering, function/tool calling, and chain-of-thought orchestration - sufficient to security-review them, not just consume them.

*Working familiarity with at least one agentic framework or orchestration toolkit (LangGraph, CrewAI, Autogen, Semantic Kernel, MCP, or comparable), and the threat surface each one introduces.

*Working knowledge of vector databases (Pinecone, Weaviate, Qdrant, pgvector) and retrieval-augmented generation patterns, with an eye to KB poisoning, retrieval leakage, and access-control failures.

*Understanding of LLM evaluation methodologies - behavioral testing, hallucination detection, output grading, drift detection - applied to security and compliance properties of agent outputs.

*Strong software engineering fundamentals in Python and/or TypeScript; able to read and contribute to production agent code, not just review it from a distance.

*Compliance fluency in SOC 2, HIPAA, GDPR, and ideally FDA 21 CFR Part 11 - and the judgment to translate those frameworks into concrete controls for AI/ML and agentic systems.

*Pharma, life sciences, or healthcare data experience (RWD/RWE, clinical trials, market access) is a strong plus given the Norstella product portfolio.

*Medical and Prescription Drug Benefits

*Health Savings Accounts (HSA) or Flexible Spending Accounts (FSA)

*Dental & Vision Benefits

*Basic Life and AD&D Benefits

*401k Retirement Plan with Company Match

*Company Paid Short & Long-Term Disability

*Paid Parental Leave

*Paid Time Off & Company Holidays

Norstella is a premier and critical global life sciences data and AI solutions provider dedicated to improving patient access to life-saving therapies. Norstella supports pharmaceutical and biotech companies across the full drug development lifecycle - from pipeline to patient. Our mission is simple: to help our clients bring therapies to market faster and more efficiently, ultimately impacting patient lives.

Norstella unites market-leading brands - Citeline, Evaluate, MMIT, Panalgo, Skipta and The Dedham Group and delivers must-have answers and insights, leveraging AI, for critical strategic, clinical, and commercial decision-making. We help our clients, Norstella serves most pharmaceutical and biotech companies around the world, along with regulators like the FDA, and payers. By providing critical proprietary data supporting AI-driven workflows, Norstella helps clients make decisions faster and with greater confidence. Norstella’s investments in AI are transforming how data is consumed and decisions are made, disrupting inefficient legacy workflows and helping the industry become more efficient, innovative, and responsive to patient needs.