Research & Publications
Academic Publications
Efficient Transformer Architectures for Edge Computing Devices
Singh, A., Zhang, L., Patel, S., & Johnson, R.
Conference on Neural Information Processing Systems (NeurIPS), 2023
This paper introduces a novel approach to optimizing transformer models for deployment on resource-constrained edge devices. We propose a combination of structured pruning, quantization-aware training, and knowledge distillation techniques that reduce model size by 85% while maintaining 96% of the original accuracy.
Distributed Training of Large Language Models: Challenges and Solutions
Singh, A., Chen, H., Williams, T., & Garcia, M.
International Conference on Machine Learning (ICML), 2022
Training large language models with billions of parameters presents significant computational challenges. This paper explores novel techniques for distributed training across GPU clusters, including pipeline parallelism, tensor parallelism, and efficient communication strategies that reduce training time by 40% compared to previous methods.
Real-time Object Detection on Embedded Systems: A Comparative Study
Singh, A., Lee, J., & Anderson, P.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022
This study evaluates the performance of various object detection algorithms on embedded platforms including NVIDIA Jetson, Raspberry Pi, and custom FPGA implementations. We provide comprehensive benchmarks and propose a novel lightweight architecture that achieves 60 FPS on Jetson devices with minimal accuracy loss.
Federated Learning for Healthcare: Privacy-Preserving Medical Image Analysis
Singh, A., Patel, S., Kim, J., & Roberts, L.
Nature Machine Intelligence, 2021
This paper presents a federated learning framework for medical image analysis that preserves patient privacy while enabling collaborative model training across multiple healthcare institutions. Our approach demonstrates how hospitals can collectively train advanced diagnostic models without sharing sensitive patient data.
Quantum Machine Learning: Current State and Future Directions
Singh, A., & Quantum Computing Research Group
ACM Quantum Computing Journal, 2023
This comprehensive review explores the intersection of quantum computing and machine learning, analyzing current quantum algorithms for ML tasks and their potential advantages over classical approaches. We discuss hardware limitations, theoretical speedups, and promising research directions for the next decade.
Conference Talks
Scaling AI Systems: From Research to Production
AI Engineering Summit, October 2023
Keynote presentation on the challenges and best practices for deploying research AI models into production environments at scale.
View PresentationThe Future of Edge AI: Opportunities and Challenges
Embedded Systems Conference, May 2023
Panel discussion on emerging trends in edge computing for AI applications and the technical hurdles that need to be overcome.
View PresentationResponsible AI Development: Ethics in Machine Learning
Tech Ethics Forum, March 2023
Talk on incorporating ethical considerations into the AI development lifecycle and mitigating algorithmic bias.
View PresentationResearch Grants
Advanced Research in Efficient Deep Learning
National Science Foundation
$1.2 Million • 2023-2026
Research grant to develop novel techniques for making deep learning models more computationally efficient and environmentally sustainable.
Edge AI for Healthcare Applications
National Institutes of Health
$850,000 • 2022-2025
Grant to develop privacy-preserving AI systems that can run locally on medical devices without transmitting sensitive patient data to the cloud.
High-Performance Computing for Climate Modeling
Department of Energy
$1.5 Million • 2021-2024
Funding for research on utilizing HPC resources to improve the accuracy and speed of climate change prediction models.