We are currently within the most dynamic and exciting era of human discovery, where the number of unanswered questions remains high and the tools required to probe them are finally being built. Among these questions, general artificial intelligence now has most of its necessary ingredients and can plausibly be developed within the next decade. It will bring opportunities for the most incredible benefits across society, yet it will likely also be the most destabilizing and dangerous technology ever created.

Within a short period of time, the majority of science needs to shift toward designing and deploying effective and safe super intelligence. This requires new neural architectures that efficiently scale while maintaining interpretability and thought-tracing, global cooperation, and targeted international regulation. The research decisions we make now will lead to a society that eventually either flourishes, decays, or even disappears.

Academics

Currently, I am a final-year PhD Candidate at the Computer Systems Laboratory at Cornell University, under the supervision of Prof. Zhiru Zhang and co-advised by Mohamed Abdelfattah. I previously studied a combination of Physics and Computer Science in the College of Arts and Sciences at Cornell University. My academic research currently focuses on building more efficient deep learning systems through efficient neural architectures, low-precision quantization, and dynamic sparsity.

Publications

• FLIQS: One-Shot Mixed-Precision Floating-Point and Integer Quantization Search
  J. Dotzel, G. Wu, A. Li, M. Umar, Y. Ni, M. S. Abdelfattah, Z. Zhang, L. Cheng, N. Jouppi, Q. Le, S. Li International Conference on Automated Machine Learning, 2024 [BEST PAPER]

• Learning from Students: Applying T-Distributions to Explore Accurate and Efficient Formats for LLMs
  J. Dotzel, Y. Chen, B. Kotb, S. Prasad, G. Wu, S. Li, M. S. Abdelfattah, Z. Zhang
  International Conference on Machine Learning, 2024

• Exploring the Limits of Semantic Image Compression at Micro-Bits per Pixel
  J. Dotzel, B. Kotb, J. Dotzel, M. S. Abdelfattah, Z. Zhang
  The Second Tiny Papers Track at ICLR, 2024

• Opportunities for Post-Training Dynamic Layer Sparsity in Large Vision and Language Models
  J. Dotzel, C. Jiang, M. Abdelfattah, Z. Zhang
  Efficient Large Vision Models Workshop at CVPR, 2024

• Semantic Compression of 3D Objects for Open and Collaborative Worlds
  J. Dotzel, T. Montes, M. S. Abdelfattah, Z. Zhang
  arXiv, 2024 (under submission)

• Radial Networks: Dynamic Layer Routing for High-Performance Large Language Models
  J. Dotzel, Y. Akhauri, A. AbouElhamayed, C. Jiang, M. Abdelfattah, Z. Zhang
  arxiv, 2024 (under submission)

• OverQ: Opportunistic Outlier Quantization for Neural Network Accelerators
  J. Dotzel*, R. Zhao*, Z. Hu, P. Ivanov, C. De Sa, Z. Zhang
  arXiv, 2019

• Improving Neural Network Quantization Without Retraining Using Outlier Channel Splitting
  R. Zhao, Y. Hu, J. Dotzel, C. De Sa, Z. Zhang
  International Conference on Machine Learning, 2019

• Building Efficient Deep Neural Networks With Unitary Group Convolutions
  R. Zhao, Y. Hu, J. Dotzel, C. D. Sa, Z. Zhang
  Conference on Computer Vision and Pattern Recognition, 2019

• ShadowLLM: Predictor-based Contextual Sparsity for Large Language Models
  Y. Akhauri, A. F. AbouElhamayed, J. Dotzel, Z. Zhang, A. M. Rush, S. Huda, M. S. Abdelfattah
  Empirical Methods in Natural Language Processing, 2024

• SparAMX: Accelerating Compressed LLMs Token Generation on AMX-powered CPUs
  A. Elhamayed, J. Dotzel, Y. Akhauri, C. Chang, …, M. Abdelfattah
  arXiv, 2024

• M4BRAM: Mixed-Precision Matrix-Matrix Multiplication in FPGA Block RAMs
  Y. Chen, J. Dotzel, M. S. Abdelfattah
  2023 International Conference on Field Programmable Technology, 2023

• Logic Synthesis Meets Machine Learning: Trading Exactness for Generalization
  S. Rai, W. L. Neto, …, Y. Zhou, Y. Zhang, J. Dotzel, Z. Zhang, …
  2021 Design, Automation & Test in Europe Conference & Exhibition, 2021

• Enabling Design Methodologies and Future Trends for Edge AI: Specialization and Co-Design
  C. Hao, J. Dotzel, J. Xiong, L. Benini, Z. Zhang, D. Chen
  IEEE Design & Test, 2021

Experience

• Google, TPU Performance Team
  Student Researcher
  June 2024 - Present

• Google, Platforms-Aware AutoML
  Student Researcher
  June 2022 - May 2024

• Computer Systems Laboratory, Cornell
  PhD Candidate
  August 2019 - Present

• Datto
  Software Engineer
  June 2017 - Jun 2018