Candidate Profile — Candidate #1
Position: d-Matrix — Software Engineer, Staff - Kernels (Santa Clara, CA)
Candidate Location: Not specified
Experience: 15.9 years experience
Relevant Experience & Education Highlights
Extensive expertise in ML compilers, hardware acceleration, and deep learning optimizations positions this candidate strongly for developing and productizing software kernels on next-generation AI hardware at d-Matrix.
1. Experience building software kernels for HW architectures
Demonstrated deep knowledge in mapping algorithms and computational graphs to specialized hardware through compiler stacks and acceleration frameworks.
- Developed MLIR-enabled compilation stack targeting commodity and in-house ML accelerators at a major autonomous vehicle technology company.
- Led frontend for PyTorch models to MLIR ecosystem, optimizing for heterogeneous compute platforms in latency-critical inference.
- Co-created and maintained Tensor Compute Primitives (mlir-tcp), an open-source MLIR dialect for ML programs.
- Co-maintained Torch-MLIR project providing PyTorch to MLIR compiler support.
2. Experience implementing algorithms for specialized hardware such as FPGAs, DSPs, GPUs, and AI accelerators
Hands-on work optimizing deep learning algorithms for programmable logic and AI engines aligns with hardware-software co-design needs.
- Advanced ML acceleration algorithms at an FPGA and adaptive computing company, including AI Engines for cloud and edge use cases.
- Developed TQT, a quantization technique for fixed-point inference on deep neural networks, accepted at a leading conference.
- Created GraFX, a source-to-source compiler framework automating PyTorch model optimization and deployment.
- Explored novel architectures for efficient ML workloads on programmable hardware.
3. Experience with ML compilers and algorithms, such as MLIR, LLVM, TVM, Glow
Proven track record building full-stack toolchains for ML model deployment bridges AI frameworks to underlying architectures.
- Worked at intersection of compilers and algorithms for deep learning optimizations and accelerated deployments.
- Tech led ML compiler efforts integrating state-of-the-art PyTorch models with MLIR middle-end.
- Designed software architecture for compiler frameworks enhancing model development lifecycles.
- Integrated components across ML training and deployment stacks.
Requirements & Candidate Alignment
| d-Matrix Requirement | Candidate Qualification |
|---|---|
| Education: MS in computer engineering, math, physics, or a related degree with 5+ years of industry experience or a PhD... with 1+ years | MS in Electrical Engineering from a top-tier research university with 15.9 years of industry experience |
| Strong grasp: of computer architecture, data structures, system software, and machine learning fundamentals | Strong alignment through work on hardware architectures, ML compilers, and deep learning optimizations for accelerators |
| Proficient: in C/C++ and Python development in Linux environments and using standard development tools | Proficient in Python with Jupyter Notebook; extensive compiler and framework development implying C/C++ proficiency |
| Experience implementing algorithms for specialized hardware such as FPGAs, DSPs, GPUs, and AI accelerators using libraries such as CUDA, etc. | Direct experience with FPGAs and AI Engines at an adaptive computing company; MLIR stacks for commodity and in-house accelerators |
| Experience in implementing operators commonly used in ML workloads—GEMMs, Convolutions, BLAS, SIMD operators... | Aligned through deep learning optimizations, quantization for neural networks, and PyTorch model deployments |
| Self-motivated team player with a strong sense of ownership and leadership | Staff-level Tech Lead roles demonstrating ownership in compiler frontends and framework creation |
| Preferred: Experience with ML frameworks such as TensorFlow and/or PyTorch | Hands-on with PyTorch including SOTA models, Torch-MLIR, TensorFlow, and GraFX framework |
| Preferred: Experience with ML compilers and algorithms, such as MLIR, LLVM, TVM, Glow, etc. | Extensive with MLIR, including Torch-MLIR, mlir-tcp dialect, and compilation stacks |
If you would like to discuss this candidate or other critical roles, here is a link to my calendar to schedule a call:
Schedule a CallContact:
Finding the signal in the noise since 2005