Lead Agentic AI Offensive Security Engineer

Hewlett Packard Enterprise•San Jose, TX

1d•Hybrid

About The Position

Hewlett Packard Enterprise’s Cybersecurity team is seeking a Lead Agentic AI Engineer to shape the technical direction of its AI-enabled offensive security organization. The role architects and delivers secure, reliable, production-grade agentic AI platforms to accelerate and scale cybersecurity triage, investigation, detection engineering, response, and other complex workflows. The position owns reference architecture and engineering standards for production agentic AI systems, and leads the design of reusable agent execution harnesses that separate AI reasoning and planning from deterministic execution. It covers multi-agent workflows, state and memory management, tool calling, human-in-the-loop controls, long-running processes, model-agnostic interfaces, retrieval and context services, and enterprise integrations. The role also establishes practices for evaluation, observability, reliability, and security, including least privilege, sandboxing, authorization, secrets management, guardrails, auditability, and defenses against prompt injection, tool abuse, and data leakage. The engineer leads cross-functional architecture decisions, mentors engineers, and communicates technical tradeoffs to senior leadership. The role is hybrid, with an average of two days per week required in an HPE office.

Requirements

At least 10 years of experience in software, platform, distributed-systems, or AI engineering, with demonstrated technical leadership on production systems.
Demonstrated expertise in at least three of the following areas: System Design & Architecture; Agentic Systems & Harness Engineering (required); Model-Agnostic AI & Context Engineering; Platform Engineering & Production Reliability; Evaluation, Security & AI Safety.
Proven experience architecting and operating production LLM-powered or agentic systems across their full lifecycle.
Deep expertise in distributed systems, microservices, APIs, event-driven architectures, workflow and state management, scalability, concurrency, and fault tolerance.
Strong hands-on experience with agent orchestration, planning, tool calling, delegation, memory, and long-running workflows using LangGraph, Claude Agent SDK, Semantic Kernel, AutoGen, or equivalent.
Experience with foundation-model APIs and open-weight or self-hosted models, structured generation, and function calling.
Expert Python and backend engineering skills, including REST/gRPC, messaging, Docker, Kubernetes, CI/CD, and cloud infrastructure.
Practical experience with agent evaluation, trajectory and regression testing, AI security, and production reliability.

Nice To Haves

Experience with cybersecurity automation, security operations, offensive security, or adversary simulation.
Experience with reusable agent platforms or open-weight inference.
Experience with frameworks such as OWASP Top 10 for LLM Applications, MITRE ATLAS, MITRE ATT&CK, or NIST AI RMF.

Responsibilities

Own reference architecture and engineering standards for production agentic AI systems.
Design reusable agent execution harnesses that separate AI reasoning and planning from deterministic execution.
Lead multi-agent workflows, state management, memory, tool calling, human-in-the-loop controls, and long-running processes.
Define model-agnostic interfaces for commercial, cloud-hosted, and open-weight models, including routing, fallback, and workload-based selection.
Architect retrieval, context, and memory services using approaches such as RAG, hybrid search, and vector databases.
Establish strategies for MCP servers, A2A protocols, APIs, tool registries, and enterprise integrations.
Define evaluation, observability, reliability, and security practices, including least privilege, sandboxing, authorization, secrets management, guardrails, auditability, and defenses against prompt injection, tool abuse, and data leakage.
Lead cross-functional architecture decisions, mentor engineers, and communicate technical tradeoffs to senior leadership.