Senior AI Security Engineer

DataSpring is the trusted data connector at the core of healthcare. The Senior AI Security Engineer is responsible for securing AI-driven technology, designing and enforcing security across the entire AI delivery surface, and embedding security into the build process while collaborating with engineering and data teams.

Responsibilities:

• Lead threat modeling for AI capabilities, agentic features, and integrations
• Maintain a living threat model aligned to OWASP Top 10 for LLMs, MITRE ATLAS, and NIST AI RMF, and translate findings into prioritized, actionable controls
• Own and maintain the AI security reference architecture covering model hosting, orchestration, tool use, MCP server deployments, and observability
• Establish security patterns for new capabilities before they reach production
• Evaluate foundation models for data-leakage risk and supply-chain provenance
• Harden prompt pipelines against injection, jailbreaking, and context poisoning
• Secure fine-tuning pipelines, RAG architectures, and embedding stores with appropriate access controls and poisoning detection
• Define trust boundaries, tool call authorization, privilege scoping, and human-in-the-loop escalation policies for multi-agent systems
• Ensure agents operate under least-privilege identities with revocable, short-lived credentials
• Establish review, testing, and gating processes for AI-generated code across frontend, backend, IaC, and CI/CD
• Embed security requirements and policy-as-code checks directly into Claude system prompts and project instructions used for development
• Enforce SCA/SBOM generation, dependency allowlisting, and secrets management for all LLM-driven build outputs
• Define and enforce identity patterns for both human and non-human (agent) identities
• Apply zero-trust principles to all model API calls, vector store queries, and external tool invocations
• Every request must be authenticated, authorized, and logged
• Architect OAuth/OIDC delegation patterns for agentic flows with bounded scopes and session limits
• Classify and control data flowing into LLM context windows, vector databases, and training pipelines
• Secure ETL and ingestion pipelines against poisoning and schema drift
• Enforce masking, tokenization, and access controls to prevent regulated data (PII, PHI, PCI) from appearing in model inputs, outputs, or logs
• Extend DLP coverage to LLM prompt submissions and completions
• Instrument AI infrastructure with behavioral telemetry
• Write detection rules for prompt injection attempts, anomalous agent behavior, and data exfiltration through LLMs
• Serve as SME for AI-related security incidents, including model abuse, pipeline compromise, and agentic runaway actions
• Support SOC 2, HIPAA, ISO 27001, and NIST AI RMF compliance as applied to AI systems
• Maintain control mappings, evidence, and audit documentation
• Evaluate and deploy AI-specific security tooling (LLM firewalls, guardrail frameworks, agent monitoring) and integrate with SIEM/SOAR

Requirements:

• 8+ years of progressive information security experience across architecture, application security, identity, and data domains
• Hands-on use of Claude for application development, including system prompt design, tool-use configuration, and multi-agent orchestration
• Experience with the Model Context Protocol (MCP) ecosystem and securing MCP server deployments
• Red team or adversarial AI testing experience: model evasion, adversarial examples, jailbreak research
• Background in a regulated industry with data sensitivity requirements, healthcare a plus
• Contributions to AI security open-source projects, published research, or conference presentations (DEF CON AI Village, Black Hat, NeurIPS)
• Demonstrated success securing AI/ML systems, LLM applications, or agentic AI platforms in a production environment
• Demonstrated experience building or securing applications developed substantially through LLM-assisted code generation (Claude, Copilot, Cursor, or equivalent)
• Cloud-native security experience on at least one major provider (AWS, Azure, GCP), including IAM policy design, network segmentation, and secrets management
• Demonstrated success automating adversarial testing infrastructure include jailbreak suites, prompt injection harnesses, and regression pipelines tied to model and prompt changes
• Relevant certifications: CISSP, CCSP, OSCP, AWS/Azure Security Specialty, or emerging AI security credentials
• Secure coding knowledge across languages commonly produced by AI-driven workflows: Python, TypeScript/JavaScript, SQL, and IaC (Terraform/Bicep)
• Strong command of identity protocols: OAuth 2.0, OIDC, SAML, SCIM, and their application to non-human identities
• Prompt engineering sufficient to construct and evaluate adversarial prompts for testing
• Experience with SIEM, CSPM, and runtime application security tools; ability to write detection logic and correlation rules
• Data security controls: encryption, tokenization, DLP, and secrets management (HashiCorp Vault, AWS Secrets Manager, Azure Key Vault)
• Scripting and automation skills for building internal security tooling (Python preferred)
• Custom CI/CD security gate development — policy checks written from scratch, not scanner configuration
• Vector database internals (pgvector, Pinecone, Weaviate) — access controls, audit logging, poisoning detection
• Red team automation for LLMs — adversarial prompt generation and evasion testing at scale
• Deep familiarity with AI orchestration frameworks (LangChain, LangGraph, AutoGen, Claude Agent SDK, or similar) and their security characteristics
• Engineer programmatic agent kill-switch and rollback mechanisms triggered by behavioral thresholds
• Track record delivering security architecture artifacts: threat models, reference architectures, security requirements, and control frameworks
• Build secure memory and context management layers for multi-agent systems — encryption, access scoping, and TTL enforcement