Spandan Tiwari

About Spandan Tiwari

Engineering leader with a strong track record in delivering successful AI software frameworks and products, especially in GenAI, LLMs, Computer Vision, ASR, and NLP.• I lead the strategy, and execution for AI frameworks at AMD\'s AI Group that serves a wide set of business units and partners. I lead an amazing team of engineers and researchers defining and developing AI framework-level capabilities to accelerate GenAI workloads on AMD\'s diverse and cutting-edge compute platforms, such as Ryzen AI NPUs and Instinct MI300x GPUs. We are enabling AMD\'s AI accelerators to all major ecosystems (PyTorch, ONNX/OnnxRuntime, HF) to enable fast and efficient inference for GenAI and predictive AI models.• At Meta, I led a team in PyTorch@Meta org developing the compiler and runtime stack for AI hardware accelerators (MTIA) for inference and training. We connected AI accelerators to the PyTorch ecosystem and enabled ML models to run blazingly fast on Meta\'s AI infrastructure.• At Meta, I was a Research Engineering Manager in the Surreal group in Facebook Reality Labs - Research, where I led a team that developed the framework for developing distributed AR/VR/XR application running on heterogeneous hardware and pushing the frontiers for social interactions in AR and VR through projects like, LiveMaps and Project Aria.• At Microsoft, I was one of the founding team members of the ONNX/OnnxRuntime standard and ecosystem. I led an engineering team in the AI Frameworks group in Microsoft, that helped develop the ONNX spec and infrastructure and OnnxRuntime backend to drive Microsoft\'s AI strategy for cross-framework interoperability in Azure. We developed exporters and interop components between ecosystems such as PyTorch, ONNX, and CNTK. I was also involved in the development of Cognitive Toolkit (CNTK) at Microsoft, which was one of the early AI frameworks.• Twenty years of experience in algorithm research, design and development software frameworks in the areas of AI/ML, computer vision, and image processing. • Driver of product-oriented research, product design and development.• Strong experience at tracking broad algorithm areas and trends and developing product roadmap. • Research Lead on the team awarded the prestigious U.S Army SBIR Achievement Award 2008. Specialities: Engineering Management, Computer Vision, Image Processing, Deep Learning, Deep Neural Networks, Algorithm Design, Machine Learning, Automatic Target Recognition, System Engineering, Multi-core Concurrent Programming, Parallelization, Code Optimization

Experience

1. Director of Engineering, Ai Frameworks and Applications, Amd Ai

   AMD

   Dec 2023 — Present · Bellevue, WA, US

   I lead the strategy and execution for AI frameworks at AMD\'s AI group that serves a wide set of business units and partners. My team is focused on enabling strategic framework level capabilities that are aimed at serving AMD\'s diverse and cutting-edge compute platforms, such as Ryzen AI NPUs and Instinct MI300x GPUs. Some aspects of my role are:* Architect for framework level tooling are developing tooling for GenAI model (LLMs, Stable Diffusion, ASR) optimization, including graph-level optimizers, quantization, sparsity, pruning and distillation.* Special focus on cutting-edge quantization techniques, with advanced mxScaling datatypes, and advanced quantization runtime features such as distributed execution and compilation for next generation of GenAI models.* Developing strategy and roadmap for specific compiler and runtime feature development (PyTorch, OnnxRuntime) for AI PC devices, that include next generation Ryzen NPU.* Managing and expanding a high-performance team - mentoring, coaching, and helping people grow their careers both as IC, TL, and manager.

Education

• Missouri University of Science and Technology

  PhD, Electrical Engineering

  2004 — 2007

• Missouri University of Science and Technology

  M.S, Electrical Engineering

  2001 — 2003

• Maulana Azad National Institute of Technology

  Bachelor of Engineering, Electronics and Telecommunication

  1997 — 2001

Skills

• R&D
• Image Processing
• Matlab
• Data Analysis
• Digital Image Processing
• Project Management
• Simulations
• Python
• Latex
• Computer Vision
• Image Segmentation
• C++
• Digital Signal Processing
• C
• Programming
• Signal Processing
• Algorithms
• Pattern Recognition
• Mathematics
• Machine Learning
• Product Development
• Embedded Systems