Job Description

Job Req ID: 30442

About Supermicro:

Supermicro® is a Top Tier provider of advanced server, storage, and networking solutions for Data Center, Cloud Computing, Enterprise IT, Hadoop/ Big Data, Hyperscale, HPC and IoT/Embedded customers worldwide. We are the #5 fastest growing company among the Silicon Valley Top 50 technology firms. Our unprecedented global expansion has provided us with the opportunity to offer a large number of new positions to the technology community. We seek talented, passionate, and committed engineers, technologists, and business leaders to join us.

Job Summary:

Supermicro is hiring a Principal AI Safety Engineer in Technology Enablement to help customers run agentic AI on Supermicro infrastructure in named enterprises and in GPU clouds (neoclouds).

This is a customer-facing design role, not a security operations role. You will join architecture workshops, define how agents and tenants are isolated, identified, governed, and audited, and write that control layer into the design before systems are quoted or shipped. Enterprise work is one customer, one data boundary. Neocloud work is many tenants on a shared factory. You will cover both. You will work with Supermicro solution architects, factory engineering, cloud and enterprise sales, and NVIDIA field teams.

Supermicro partners with NVIDIA, AMD, and Intel, and with leading storage and networking companies. This role is majority NVIDIA-focused, including NVIDIA runtime and guardrail software, confidential computing, and data-center security features where they belong in the design. The same control model must still work when a factory includes AMD, Intel, or partner storage and networking.

San Jose, CA location is preferred. Exceptional Principal-level candidates elsewhere in the United States will be considered if they can travel for customer workshops.

Essential Duties and Responsibilities:

1. Customer-facing AI safety architecture — enterprise and neocloud

Join engagements when customers will deploy agents, tools, or regulated inference.

• Enterprise: safety profile for a single owner — data boundary, air-gap or private cloud, human gates, audit, and what OpenShell vs Sentry vs confidential computing actually enforce in that boundary.
• Neocloud: tenant isolation, inference-gateway policy, quota and blast-radius controls, confidential computing and DPU isolation between tenants, and attested evidence a cloud customer can hand to their own CISO (identity, policy decisions, telemetry — not agent self-reporting).
• Set proof-of-concept pass/fail criteria so safety is tested, not assumed. Define what is enforceable on current installed NVIDIA systems versus what requires BlueField-4 Sentry or Vera-class platforms.

1. Controls on Supermicro AI factories

Map the NVIDIA Open Agent Safety Platform onto Supermicro server, rack, and cluster designs used by enterprises and GPU cloud operators.

• Place OpenShell (runtime, least-privilege policy, sandbox) and Sentry (BlueField-4 / DOCA, out-of-band monitoring, millisecond containment) in the design.
• Identify where confidential computing, SmartNIC/DPU security, multi-tenant networking, air-gapped operation, or partner security platforms belong.
• Design one control model that still works when a factory includes AMD, Intel, or partner storage and networking. OpenShell’s extendability off NVIDIA CPUs is part of that model, not an afterthought.
• Produce the safety chapter of the quote package: what is in hardware, what is in runtime, what remains a customer operational control.

1. Field enablement and demonstration

Create safety design templates for both motions: named-enterprise factory and multi-tenant neocloud. Help solution architects start the conversation; you own the safety architecture when agents or tenancy are in scope.

Contribute customer-facing material and represent Supermicro in technical sessions with enterprises, GPU cloud operators, NVIDIA, and other partners. You will put a working control demo in front of a customer and an NVIDIA architect. OpenShell at minimum; Sentry / BlueField-4 / confidential computing where the platform supports it.

Qualifications:

• Bachelor’s or master’s in Computer Science, Computer Engineering, Information Technology, or equivalent.
• 12+ years in security engineering, platform security, cloud security, or applied AI systems. 5+ years accepted with hands-on production work on agent or LLM runtime controls (sandbox, policy engine, gateway, or equivalent — not training-only or policy-paper backgrounds).
• Able to prototype and demonstrate controls: Python or equivalent, containers, and policy or guardrail configuration.
• Hands-on with at least two of: NVIDIA OpenShell, NVIDIA Sentry / BlueField / DOCA, confidential computing, inference gateway, NIM, or NeMo Guardrails. OpenShell or Sentry strongly preferred after this announcement.
• Hands-on experience with sandboxes, policy engines, secret proxies, identity for agents, or equivalent.
• Experience with at least one of: regulated enterprise AI or multi-tenant GPU / AI cloud isolation. Both strongly preferred.
• Working knowledge of NVIDIA enterprise and safety software and how it lands on server, DPU, and rack designs. NVIDIA certification desirable.
• Ability to design controls for mixed GPU/CPU environments and for shared-cluster tenancy.
• Comfortable presenting to customer security leaders, cloud architects, and NVIDIA technical teams.

Salary Range

​$210,000 - $255,000

The salary offered will depend on several factors, including your location, level, education, training, specific skills, years of experience, and comparison to other employees already in this role. In addition to a comprehensive benefits package, candidates may be eligible for other forms of compensation, such as participation in bonus and equity award programs.

​

EEO Statement

Supermicro is an Equal Opportunity Employer and embraces diversity in our employee population. It is the policy of Supermicro to provide equal opportunity to all qualified applicants and employees without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, protected veteran status or special disabled veteran, marital status, pregnancy, genetic information, or any other legally protected status.