Interested in machine learning, optimization, deep neural networks, data science, and a lot of other things! I've worked on mathematical modeling and optimization in domains such as language models, computer vision, physiology, and cell biology.

Workplaces

Employer	Role	Dates
NVIDIA	Deep Learning Algorithm Engineer	2024-
Samsung Semiconductor, Inc.	Deep Learning Engineer / Research Scientist	2019-2024
Mythic AI	Software Engineer (Neural Network Co-Design)	2017-2019
Quantcast	Software Engineer	2016-2017
Teradata Aster	Analytics Engineer	2014-2016
Immunetrics	Mathematical Modeling	2013-2014
Henry M. Jackson Foundation	Research Scientist	2010-2013
University of Washington	Postdoctoral Scholar	2006-2010

Download a full resume, if you wish.

Publications

AI, Robotics & Machine Learning

• S. Kim*, S. Shen*, D. Thorsley*, A. Gholami, W. Kwon, J. Hassoun, and K. Keutzer, Learned Token Pruning for Transformers. In KDD '22: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, August 2022.

  • Related Tech Report from UC-Berkeley

• J. Fang, A. Shafiee, H. Abdel-Aziz, D. Thorsley, G. Georgiadis, and J.H. Hassoun, Post-training Piecewise Linear Quantization for Deep Neural Networks. In Proceedings on the 2020 European Conference on Computer Vision, August 2020.

  • Related patent pending

• R. Han, M. Zheng, S. Byna, H. Tang, B. Dong, D. Dai, Y. Chen, D. Kim, J. Hassoun, D. Thorsley, and M. Wolf. PROV-IO+: A Cross-Platform Provenance Framework for Scientific Data on HPC Systems. IEEE Transactions on Parallel and Distributed Systems, 35 (5), pp. 844-861, May 2024.

• N. Napp, D. Thorsley, and E. Klavins, Hidden Markov Models for non-Well-Mixed Reaction Networks. In Proceedings of the 2009 American Control Conference, pp. 737-744, June 2009.

Unpublished (available online)

• L. Li, D. Thorsley and J. Hassoun. SaiT: Sparse Vision Transformers through Adaptive Token Pruning. Available on arXiv, October 2022.

• J. Fang, L. Yang, C. Shen, H. Abdel-Aziz, D. Thorsley and J. Hassoun. Fast and Efficient Once-For-All Networks for Diverse Hardware Deployment. Available on OpenReview, 2021.

Control Theory & Diagnosis

• D. Thorsley. A necessary and sufficient condition for diagnosability of stochastic discrete event systems. Discrete Event Dynamic Systems: Theory and Applications, 27 (3), pp. 481–500, September 2017.

• D. Thorsley. Diagnosability of Stochastic Chemical Kinetic Systems: A Discrete Event Systems Approach. In Proceedings of the 2010 American Control Conference, pp. 2623-2630, June 2010.

• W.-C. Lin, H.E. Garcia, D. Thorsley, and T.-S. Yoo, Sequential Window Diagnoser for Discrete-Event Systems under Unreliable Observations. In Proceedings of the 47th Allerton Conference on Communication, Control, and Computing, pp. 668-675, October 2009.

• D. Thorsley, T.-S. Yoo, and H.E. Garcia. Diagnosability of Stochastic Discrete-Event Systems Under Unreliable Observations. In Proceedings of the 2008 American Control Conference, pp. 1158-1165, June 2008.

• D. Thorsley and D. Teneketzis. Active Acquisition of Information for Diagnosis and Control of Discrete-Event Systems. Discrete Event Dynamic Systems: Theory and Applications, 17 (4), pp. 531–583, December 2007.

• D. Thorsley and D. Teneketzis. Actuator Failure in Decentralized Supervisory Control Systems. In Proceedings of the 1st IFAC Workshop on Dependable Control of Discrete Systems, pp. 17–22, June 2007.

• D. Thorsley and D. Teneketzis, Intrusion Detection in Controlled Discrete Event Systems. In Proceedings of the 45th IEEE Conference on Decision and Control, pp. 6047-6054, Dec. 2006.

  • Note: the version of this paper on IEEE Xplore has corrupted symbols. This version is the correct one.

• D. Thorsley and D. Teneketzis. Diagnosis of Cyclic Discrete-Event Systems Using Active Acquisition of Information. In Proceedings of the 8th International Workshop on Discrete Event Systems (WODES ’06), pp. 248– 255, July 2006.

• D. Thorsley and D. Teneketzis. Diagnosability of Stochastic Discrete-Event Systems. IEEE Transactions on Automatic Control, 50 (4), pp. 476–492, April 2005.

• D. Thorsley and D. Teneketzis. Active Acquisition of Information for Diagnosis of Discrete Event Systems. In Proceedings of the 42nd Allerton Conference on Communication, Control, and Computing, pp. 562-571, September 2004.

• D. Thorsley and D. Teneketzis. Diagnosability of Stochastic Automata. In Proceedings of the 42nd IEEE Conference on Decision and Control, pp. 6289–6294, December 2003.

• D. Thorsley and D. Teneketzis. “Failure Diagnosis of Stochastic Automata.” In Proceedings of the 14th International Workshop on the Principles of Diagnosis (DX-03), pp. 131–137, June 2003.

Synthetic Biology & Physiology

• M.Y. Khitrov, S. Laxminarayan, D. Thorsley, S. Ramakrishnan, S. Rajaraman, N.J. Wesensten, et al., PC-PVT: A Platform for Psychomotor Vigilance Task Testing, Analysis, and Prediction Behavioral Research Methods 46 (1), pp. 140-147, March 2014.

• P. Rajdev, D. Thorsley, S. Rajaraman, T.L. Rupp, N.J. Wesensten, T.J. Balkin, et al., A Unified Mathematical Model to Quantify Performance Impairment for Both Chronic Sleep Restriction and Total Sleep Deprivation. Journal of Theoretical Biology, 331 (1) pp. 66-77, August 2013.

• S. Rajaraman, S. Ramakrishnan, D. Thorsley, T.L. Rupp, N.J. Wesensten, T.J. Balkin, et al., A New Metric for Quantifying Performance Impairment on the Psychomotor Vigilance Test. Journal of Sleep Research, 21 (6) pp. 659-674, December 2012.

• D. Thorsley, R. Leproult, K. Spiegel, et al., A Phenomenological Model for Circadian and Sleep Allostatic Modulation of Plasma Cortisol Concentration. American Journal of Physiology – Endocrinology and Metabolism, 303 (10), pp. E1190-E1201, November 15, 2012.

• D. Thorsley and E. Klavins, Estimation and Discrimination of Stochastic Biochemical Circuits From Time-Lapse Microscopy Data. PLOS ONE, November 6, 2012. doi:10.1371/journal.pone.0047151. Featured in PLOS Collections: Synthetic Biology.

• S. Ramakrishnan, S. Laxminarayan, D. Thorsley, N.J. Wesensten, T.J. Balkin, et al., Individualized Performance Prediction During Total Sleep Deprivation: Accounting for Trait Vulnerability to Sleep Loss In Proceedings of the 34th Annual Conference of the IEEE Engineering in Medicine & Biology Society, pp. 5574-5577, August 2012.

• D. Thorsley and E. Klavins, Approximating Stochastic Biochemical Processes With Wasserstein Pseudometrics. IET Systems Biology, 4 (3), pp. 193-211, May 2010.

• D. Thorsley and E. Klavins, “A Theory of Approximation for Stochastic Biochemical Networks,” P. Iglesias and B. Ingalls, eds., Control Theory and Systems Biology, MIT Press, pp. 243-264, 2009.

• D. Thorsley and E. Klavins. Model Reduction of Stochastic Processes Using Wasserstein Pseudometrics. In Proceedings of the 2008 American Control Conference, pp. 1374-1381, June 2008.

Patents and Disclosures

• Systems and methods for matrix operation selector based on machine learning

• Creating an accurate latency lookup table for NPU

• Piecewise Quantization for Neural Networks

• Learned threshold token pruning for transformer neural networks

• System and method for training a neural network under performance and hardware constraints

• Method for sparsification of feature maps in self-attention mechanisms

• Shrink DARTS