Responsibilities:
Architectural Red Teaming: Design and execute targeted adversarial campaigns against the hardware/software interface of modern SoCs, specifically focusing on CPU and GPU.
Vulnerability Discovery: Conduct deep-dive vulnerability research into microarchitectural behaviors, side-channels, and hardware-assisted security features.
Exploit Development: Develop proof-of-concept exploit chains that bridge hardware vulnerabilities into kernel space, demonstrating real-world impact.
GPU & Compute Security: Analyze GPU driver interfaces, command processors, and execution units, particularly in the context of modern compute workloads and hardware acceleration.
Mitigation Bypass & Analysis: Evaluate and attempt to bypass modern hardware mitigations (e.g., Pointer Authentication, Memory Tagging, virtualization boundaries).
Cross-Functional Collaboration: Partner with hardware engineering and software kernel teams to translate your red team findings into robust, architectural-level mitigations.
Minimum Qualifications:
Deep Architectural Knowledge: Strong understanding of modern CPU architectures (pipeline stages, caching mechanisms, MMU, speculative execution). Extensive familiarity with ARM v8/v9 or x86 instruction sets.
GPU/Compute Expertise: Proven understanding of GPU architectures (execution pipelines, memory hierarchies) and how software interfaces with them (e.g., driver vulnerability research, compute batching, memory isolation).
Low-Level Software Mastery: Proficiency in C, C++, and Assembly. Deep understanding of operating system internals, kernel architecture, and memory management.
Proven Track Record: Demonstrable experience in discovering vulnerabilities or developing exploits at the OS/Hypervisor level, or in hardware-software interfaces (CVEs, bug bounties, conference talks, or whitepapers).
Preferred Qualifications:
Experience with hardware description languages (Verilog/VHDL) or pre-silicon security validation.
Experience with confidential computing environments.
Familiarity with the security implications of AI/ML hardware accelerators and complex GPU workloads (e.g., memory handling in PagedAttention or Mixture of Experts).
Familiarity with reverse engineering tools (IDA Pro, Ghidra) and dynamic analysis/debugging tools.
