Senior Technical Fellow, Parallel Programming Performance Optimization & GPU Software Diagnostics

Advanced Micro Devices, Inc•Austin, TX

About The Position

At AMD, our mission is to build great products that accelerate next-generation computing experiences—from AI and data centers, to PCs, gaming and embedded systems. Grounded in a culture of innovation and collaboration, we believe real progress comes from bold ideas, human ingenuity and a shared passion to create something extraordinary. When you join AMD, you’ll discover the real differentiator is our culture. We push the limits of innovation to solve the world’s most important challenges—striving for execution excellence, while being direct, humble, collaborative, and inclusive of diverse perspectives. Join us as we shape the future of AI and beyond. Together, we advance your career. The Senior Technical Fellow role is a company-level technical leadership position, accountable for defining and driving AMD’s strategy, architecture, and tooling to achieve industry-leading GPU performance and reliability. You will partner across hardware architecture, silicon design, firmware, drivers, compilers, runtime, and developer tools to architect scalable performance methodologies, build advanced diagnostics to detect and isolate defects, and ensure product excellence from pre-silicon through customer deployment.

Requirements

GPU/accelerator or HPC performance engineering, compilers, runtime systems, and/or reliability/diagnostics.
Expert-level proficiency in parallel programming models such as (HIP/ROCm, OpenCL, SYCL; CUDA interop), and performance tuning at scale.
Deep knowledge of GPU microarchitecture (compute units/SIMD, wavefront scheduling, memory hierarchy, caches, shared/local memory, interconnects, PCIe/Infinity Fabric).
Hands-on experience with profiling, instrumentation, and performance counters; familiarity with tools such as ROCm tools, Radeon GPU Profiler/Memory Visualizer, and LLVM-based toolchains.
Proven track record building diagnostics: sanitizers, static/dynamic analysis, fuzzing, crash triage systems, telemetry pipelines; experience with kernel-level debugging and driver/runtime fault isolation.
Strong systems background: Linux, kernel, drivers, multi-node HPC/cloud, NUMA, distributed training/inference pipelines.
Proficiency in C/C++, Python; familiarity with LLVM IR, GPU ISA, and performance modeling.
Demonstrated impact influencing architecture and shipping high-performance, reliable products at scale.

Nice To Haves

Experience with AMD GPU architectures (CDNA/RDNA) and ROCm ecosystem.
Contributions to open-source compilers/runtimes/tools; leadership in standards bodies.
Experience with ML/DL compiler stacks (TVM, XLA, Triton) and performance engineering for AI workloads.
Background in RAS, fault tolerance, error containment, and predictive analytics for accelerators.

Responsibilities

Technical Strategy and Leadership
Define the long-term technical roadmap for GPU performance engineering, diagnostics, and reliability/availability/serviceability (RAS).
Establish best-in-class methodologies for performance modeling, profiling, and optimization across ROCm/HIP, OpenCL, SYCL, CUDA interoperability, and ML/DL frameworks.
Influence architecture decisions with data-backed insights on compute, memory hierarchy, interconnect, scheduling, and compiler/runtime impacts.
Parallel Programming Performance Optimization
Architect end-to-end performance workflows: microbenchmarks, workload characterization, bottleneck analysis, instrumentation, and guided optimization.
Lead development of profiling and visualization capabilities (e.g., pipeline stages, wavefront occupancy, cache/memory behavior, interconnect utilization, synchronization overhead).
Drive compiler and runtime optimizations, including code generation, vectorization, register allocation, memory tiling, and kernel fusion.
Build scalable auto-tuning systems for kernels (GEMM/conv, attention, graph workloads) across different GPU generations and system topologies.
Software Diagnostics and Defect Detection
Design advanced diagnostics to detect, localize, and triage GPU defects across silicon, firmware, driver, runtime, and application layers.
Develop static/dynamic analysis tools, sanitizers, fuzzing, and differential testing frameworks targeted at GPU execution and memory models.
Integrate telemetry and performance counters for fault detection (ECC, parity, timeouts, hangs, deadlocks, memory corruption, race conditions) and create automated triage pipelines.
Collaborate on DFT/DFD strategies, bring-up, validation, and manufacturing test enhancements to improve yield and field reliability.
Partner with RAS teams to enhance error containment, recovery, and predictive failure analytics; contribute to BIST and in-field diagnostics strategy.
Cross-Functional Collaboration
Work closely with Architecture, Silicon Design, Firmware, Drivers, Compiler/Runtime, Tools, QA/Validation, and Customer Engineering to deliver performance and reliability targets.
Serve as a primary technical interface for strategic customers, ISVs, hyperscalers, and HPC partners, shaping workload optimization and diagnostics deployment.
Mentor senior/principal engineers and lead technical reviews; cultivate a culture of rigorous measurement, reproducibility, and engineering excellence.
External Leadership and Thought Influence
Publish influential research, secure patents, and present at top-tier conferences (SC, Hot Chips, ISCA, PACT, MICRO, GTC).
Contribute to standards and open-source initiatives in heterogeneous computing, performance tools, and reliability.