Kushagra Pundeer

Biography

Kushagra Pundeer is an AI/ML Algorithm Engineer in the Platform and ASIC Research (PAR) Lab at Nokia Bell Labs, where he leverages hardware-software co-design techniques to enable efficient deployment of AI and machine learning models on next-generation ASIC and edge computing platforms. His current research focuses on optimizing foundation models for custom compute-in-memory (CIM) ASIC architectures, enabling energy-efficient execution of AI workloads under stringent performance, memory, and power constraints. His work spans AI kernel development, architecture exploration, and simulation-driven design frameworks that guide hardware-software co-design decisions for emerging AI accelerators.
With more than five years of experience in the semiconductor industry, Kushagra has worked at the intersection of artificial intelligence, computer vision, and specialized computing platforms. His expertise includes optimization of AI and machine learning models for ASICs and FPGAs, hardware-aware machine learning, edge AI, generative AI, large language models, and computer vision systems. At Nokia Bell Labs, he contributes to research on optimizing foundation models such as LLaMA and ResNet-class architectures on novel accelerator platforms, leveraging software simulation, benchmarking, and kernel-level optimization to maximize performance and energy efficiency. His work also involves evaluating and optimizing AI workloads against industry-standard benchmarks, including MLPerf and publicly available large language model benchmarks. Prior to joining Nokia Bell Labs, he spent more than four years at AMD/Xilinx as a Machine Learning Engineer in the Video Smart Streaming team, where he developed machine learning solutions for video conferencing and live video streaming applications, improving visual quality while reducing bandwidth requirements through FPGA-accelerated processing. He also contributed to SDK development and deployment frameworks that enabled efficient AI-powered video optimization across cloud and edge environments.
Kushagra holds a Master of Science in Computer Science from the University of Massachusetts Amherst, where he specialized in Machine Learning, Computer Vision, and Robotics, and a bachelor’s degree in mechanical engineering and computer science from Drexel University. In addition to his industrial experience, he has worked on collaborative research projects with industry partners, including initiatives associated with the Chan Zuckerberg Initiative and Voya Investment Management, producing research-oriented outcomes that bridged academic innovation with real-world applications. He has participated in conferences and technical forums focused on emerging topics such as Quantum Technologies and Machine Learning and has authored numerous technical research and project papers throughout his academic and professional career. His broader research interests include large language models, energy-efficient AI infrastructure, edge intelligence, computer vision, and AI accelerator architectures, with a particular focus on using simulation, benchmarking, and hardware-software co-design to bridge the gap between algorithmic innovation and practical, scalable AI systems.