The posting
Description:
- Lead applied research and strategic definition of machine learning algorithms, quantization methodologies, and toolchain capabilities for the Neural Network Development Kit (NDK) roadmap targeting next-generation Edge AI compute solutions
- Drive innovation at the intersection of ML algorithms and constrained hardware environments, identifying and validating the latest Edge AI technologies applicable to product requirements
- Serve as the primary technical interface between the AI Architecture team and the Software (Edge AI) team, delivering well-researched toolchain feature proposals and algorithmic specifications for implementation
- Collaborate with Software (Edge AI) and AI Architecture teams to identify and pursue targeted improvements in ML software methodology, and support Software-initiated improvement efforts with algorithmic insight and implementation guidance
- Maintain deep engagement with the global Edge AI research community to ensure the NDK roadmap reflects the state of the art in model efficiency, compression, and on-device learning
Key Responsibilities:
- Collaborate with the Sr. Design Manager (AI Architect) to define and maintain the NDK toolchain feature roadmap, ensuring alignment with the NPU hardware roadmap and overall AI product strategy
- Research, evaluate, and recommend quantization algorithms, pruning strategies, knowledge distillation techniques, and other model compression methodologies suited to constrained hardware targets
- Assemble and lead focused task forces drawing on partial bandwidth from the Software (Edge AI) and IC Design teams to prototype, benchmark, and validate proposed toolchain concepts before broader commitment
- Prototype and benchmark candidate ML algorithms and toolchain features to quantitatively demonstrate accuracy-performance trade-offs and justify roadmap prioritization
- Translate hardware architectural capabilities and constraints (as defined by the NPU Architect) into concrete toolchain feature requirements and algorithmic optimization opportunities
- Deliver comprehensive technical specifications and algorithmic documentation to the Software (Edge AI) team to enable confident and accurate implementation of NDK features
- Collaborate closely with the Software (Edge AI) team throughout the implementation phase to resolve algorithmic questions, validate correctness of implementations, and ensure performance targets are met
- Actively monitor and synthesize developments from the Edge AI research community — including publications, open-source frameworks, and industry benchmarks — to continuously inform and refresh the NDK roadmap
- Partner with the Software (Edge AI) team to jointly identify ML toolchain methodology improvement opportunities and drive those that originate from the AI Architecture team; provide expert advisory support for methodology improvements initiated by the Software team
- Evaluate and apply a range of productivity tools and techniques — including but not limited to AI-assisted methods — to accelerate algorithmic prototyping, benchmarking, and specification productivity
- Evaluate and integrate relevant open-source ML frameworks, runtimes, and toolchain components (e.g., MLIR, TVM, ONNX Runtime) as acceleration vectors for NDK development
Requirements:
- Bachelor's or Master's degree in Computer Science, Electrical Engineering, Computer Engineering, or related technical field; PhD preferred, particularly in machine learning, optimization, or computer architecture
- 8+ years of experience in applied machine learning engineering or Edge AI software, with at least 5 years focused on model optimization, ML compilers, or on-device inference toolchain development
- Proven expertise in quantization (PTQ, QAT), pruning, knowledge distillation, and other model compression techniques with demonstrated results on resource-constrained hardware
- Strong knowledge of AI/ML algorithms, neural network architectures (CNNs, RNNs, Transformers, etc.), and the trade-offs between model accuracy, computational complexity, and memory footprint
- Demonstrated ability to stay at the forefront of the Edge AI research community, with a track record of translating academic and industry advances into practical product roadmap contributions
- Hands-on experience with mainstream ML frameworks (PyTorch, TensorFlow/Lite) and familiarity with ML compiler stacks such as MLIR, TVM, or ONNX Runtime
- Experience consuming hardware architectural specifications and translating them into software toolchain requirements and algorithmic optimizations
- Excellent communication skills with ability to present complex research findings and toolchain proposals clearly to architecture, software, and executive audiences
- Strong analytical and problem-solving abilities with emphasis on quantitative benchmarking, accuracy-efficiency trade-off analysis, and performance profiling on target hardware
- Demonstratedability to work collaboratively across team boundaries, including assembling and coordinating cross-functional task forces without direct authority
- Familiarity with RISC-V ISA and its software ecosystem, particularly in the context of AI inference deployment
- Experience with FPGA-based or simulator-based prototyping to validate algorithmic concepts against pre-silicon hardware models (preferred but not required)
- Self-motivated with ability to work independently, lead applied research initiatives, and drive toolchain innovation from algorithmic exploration through specification and successful team handoff
