Software Security Researcher, GPU and AI Systems

WHAT YOU WILL BE WORKING ON

You will own security research into the GPU and AI infrastructure stack, and turn what you find into product.

This means understanding how NVIDIA GPUs and adjacent accelerator platforms behave under attack: the driver and runtime layers, container and virtualization boundaries, GPU sharing and multi-tenancy models, model and dependency supply chains, and the workload patterns of real training and inference systems.

Your research directly shapes what the Stealthium platform detects. You will not be writing findings that sit in a document. You will be defining the threat models, building the proof of concepts, and working alongside engineering to convert them into runtime detections that ship.

This is a hands-on research role. Expect to spend your time in drivers, kernels, container runtimes, firmware interfaces, and real GPU clusters.

WHAT YOU WILL DO

• Conduct applied security research across GPU platforms, accelerator runtimes, and AI infrastructure components
• Build proof-of-concept exploits and attack simulations that demonstrate real risk in GPU and AI workloads, then translate them into detection logic
• Perform threat modeling for AI compute environments: multi-tenant GPU isolation, model and container supply chains, driver and firmware attack surface, orchestration and scheduling layers
• Reverse engineer components of the GPU and AI software stack to understand undocumented behavior and identify weaknesses
• Partner with engineering to define what the platform should observe, why it matters, and how an attack looks in telemetry
• Contribute to core product capabilities around runtime insight, anomaly detection, and supply chain integrity
• Track the external threat landscape for AI infrastructure and bring that intelligence back into the roadmap
• Represent Stealthium's research externally through publications, disclosures, and conference work

WHAT WE ARE LOOKING FOR

• 3+ years in security research, vulnerability research, exploit development, or offensive security engineering
• Degree in computer science, electrical engineering, or a related field, or equivalent hands-on experience
• Strong systems fundamentals: Linux internals, memory management, kernel and userspace boundaries, process and container isolation
• Practical experience with at least one of: driver or kernel research, firmware security, virtualization and container escape research, or hardware-adjacent attack surface
• Comfort reading and writing systems code, in C, C++, Rust, Go, or Python, and building your own tooling
• Working knowledge of applied cryptography, common attack vectors, and mitigation techniques
• Ability to work independently in an early-stage environment, take a vague problem, and return with something concrete

NICE TO HAVE

• Direct experience with NVIDIA GPU internals: CUDA, drivers, MIG, vGPU, GPUDirect, DOCA, or BlueField
• Familiarity with AI and ML infrastructure: training and inference pipelines, model formats and serialization risks, ML dependency supply chains
• Experience with Kubernetes, container runtimes, and GPU scheduling in production
• Cloud security experience, especially in multi-tenant or shared infrastructure environments
• Background in observability or detection engineering: metrics, logs, traces, profiling, eBPF
• Published research, CVEs, conference talks, or open source contributions
• M.Sc. or Ph.D. in computer science, electrical engineering, or a related technical field
• Experience with high-performance computing environments or large-scale GPU deployments