Peetak Mitra, Ph.D.
I am a fifth (and final) year Ph.D. candidate at the University of Massachusetts, Amherst. I work in the area of Machine Learning and predictive fluid modeling and I am advised by David Paul Schmidt. I completed part of my dissertation research at the Los Alamos National Laboratory, where I worked on developing machine learning frameworks to model geophysical turbulence processes, a critical driver of energy exchange in the oceans and atmosphere. I spent the summer of 2020 working (remotely) with TOTAL S.A. R&D, in developing High Performance Machine Learning codes using TensorFlow and PyTorch.
The other major thrust of my research is within the ICEnet Consortium, an industry-funded data-driven consortium that I co-founded with my doctoral advisor Prof. David P. Schmidt. ICEnet leverages the power of Artificial Intelligence (AI) to develop predictive machine learning algorithms to improve engine design, helping create fuel-efficient engines capable of meeting stricter environmental regulations. Among the partners for ICEnet include leading engine CFD software makers/users such as SIEMENS-CD/Adapco, Cummins, Convergent Science and AVL as well as NVIDIA and MathWorks.
For my early Ph.D. research I worked on developing reduced order models (ROM) for advanced propulsion systems, working in close collaboration with U.S. Department of Energy national laboratories - Argonne National Laboratory and Sandia National Laboratories, in a $2.1 million industry funded project, Spray Combustion Consortium. Models that I have developed have been implemented within commercial software code Converge CFD, and is in use by major automotive engine makers from around the world.
Apart from my research, I am a part of the core-team at Climate Change AI, a volunteer network that catalyzes impactful work at the intersection of machine learning and climate change and lead the monthly newsletter. I have regularly served as a reviewer for prestigious machine learning journals and conferences such as Nat. Machine Intelligence, and NeurIPS, among others.
Upcoming invited talk at 16th U.S. National Congress on Computational Mechanics, 7/21. Link: TBP
Talk on using Machine Learning for fluids accepted at NVIDIA-GTC 2021, will be available 4/21. Link: TBP
Presented talk at SIAM CSE 2021: Learning Operators from Data Symposium 2021, Dallas, TX, 'Non-Intrusive Machine Learning Models for Fluid Simulations'
Paper accepted to SAE WCX 2021: 'Analysis and Interpretation of Data driven closure models relevant to Internal Combustion Engines'
Presented three papers at Neural Informational Processing Systems workshops
Blog featured by Towards Data Science on Medium [Link]
Blog published by MathWorks [Link]
Blog published by Venkat Viswanathan [Carnegie Mellon University] [Link]
Presented three talks at the 73rd meeting of the American Physical Society's Division of Fluid Dynamics 2020 meeting
As part of my community service, I advise undergraduate and early graduate degree (first year Masters) students. Additionally I regularly review papers from prestigious scientific journals and machine learning conferences, including:
Nature Machine Intelligence Journal
Neural Networks Journal
Atomization and Sprays Journal
Neural Information Processing Systems (NeurIPS) Workshops on:
Machine Learning for the Physical Sciences 2020,
Algorithmic Fairness through the lens of causality and Interpretability, 2020
Tackling Climate Change with Machine Learning workshop at the International Conference on Learning Representations (ICLR) 2020 and a meta-reviewer for the International Conference on Machine Learning (ICML) 2021
Papers and Presentations :
Haghshenas, M., Mitra, P., Wang, C., Tagliante, F., Senecal, P.K., Pickett, L.P., Schmidt, D.P., Improved Methods for Mixing-Limited Spray Modeling, submitted to ILASS 2021
Mitra, P.P., Dal Santo, N., Haghshenas, M., Mitra, S.P., Schmidt D.P., Pruning strategies for Scientific data, submitted to ICLR 2021 Workshop
Mitra, P.P., Schmidt D.P., Characterizing cavitation within single-hole cylindrical and multi-hole convergent nozzles, to be submitted to Atomization and Sprays Journal
Presented (including submitted) :
Schmidt D.P., Mitra, P. P., Haghshenas, M., Wang, C., Tagliante, F., Senecal, P.K, Pickett, L.M., Eulerian Lagrangian Mixing Oriented (ELMO) Model, submitted to International Journal of Multiphase Flow
Mitra, P.P., Haghshenas, M., Dal Santo, N., Mitra, S.P., Schmidt, D.P., Machine Learning for Internal Combustion Engine Simulations, submitted to SAE WCX 2021
Haghshenas, M., Mitra, P.P., Dal Santo, N., Dias Ribeiro, M., Mitra, S.P., Schmidt, D.P., LES Turbulence Model with Learnt Closure; Integration of DNN into a CFD Solver, to feature at American Physical Society DFD 2020
Mitra P.P., Haghshenas, M., Dal Santo, N., Daly, C., Mitra, S.P., Schmidt, D.P., Towards Building Robust Neural Network Models for Fluid Simulations, to feature at American Physical Society DFD 2020
Mitra, P.P., Haghshenas, M., Schmidt, D.P., Super Resolution approach for accelerating fluid solvers, to feature at American Physical Society DFD 2020
Shankar, V., Portwood, G.D., Mohan, A.T., Mitra, P.P., Viswanathan, V., Schmidt, D.P., Rapid Spatiotemporal Turbulence Modeling with Convolutional Neural ODEs, to feature at American Physical Society DFD 2020
Tekawade, A., P. Mitra, B. A. Sforzo, K. E. Matusik, A. L. Kastengren, D. P. Schmidt, and C. F. Powell. "A comparison between CFD and 3D X-ray Diagnostics of Internal Flow in a Cavitating Diesel Injector Nozzle." In ILASS-Americas 30th Annual Conference on Liquid Atomization and Spray Systems, Tempe, AZ. 2019.
Ribeiro, Mateus Dias, Gavin D. Portwood, Peetak Mitra, Tan Mihn Nyugen, Balasubramanya T. Nadiga, Michael Chertkov, Anima Anandkumar, and David P. Schmidt. "A data-driven approach to modeling turbulent decay at non-asymptotic reynolds numbers." Bulletin of the American Physical Society (2019).
Portwood, Gavin D., Peetak P. Mitra, Mateus Dias Ribeiro, Tan Minh Nguyen, Balasubramanya T. Nadiga, Juan A. Saenz, Michael Chertkov et al. "Turbulence forecasting via Neural ODE." arXiv:1911.05180 (2019).
Mitra, Peetak, Mateus Dias Ribeiro, and David Schmidt. "A data-driven approach to modeling turbulent flows in an engine environment." APS (2019): G16-003.
Mitra, Peetak, and Suhrid Deshmukh. "Modeling PKT at a global level: A machine learning approach." arXiv preprint arXiv:1908.00624 (2019).
Mitra, Peetak, Katarzyna Matusik, Daniel Duke, Priyesh Srivastava, Koji Yasutomi, Julien Manin, Lyle Pickett et al. Identification and characterization of steady spray conditions in convergent, single-hole diesel injectors. No. 2019-01-0281. SAE Technical Paper, 2019.
Mitra, Peetak. "Pedestrian Collision Avoidance System (PeCAS): a Deep Learning Framework." arXiv preprint arXiv:1811.04453 (2018).
Colombo, Alessandro, Pierangelo Conti, Maurizio Orlandi, Federico Visconti, Peetak Mitra, and David P. Schmidt. "CFD simulations of a two-phase ejector for transcritical CO2 cycles applied to supermarket refrigeration systems." In 13th IIR Gustav Lorentzen Conference on Natural Refrigerants: Natural Refrigerant Solutions for Warm Climate Countries, pp. 403-410. International Institute of Refrigeration, 2018.
Noyes, Matthew A., Peetak Mitra, and Antariksh Dicholkar. "Propagation of Surface-to-Low Earth Orbit Vortex Rings for Orbital Debris Management." ESASP 715 (2013): 11.
Noyes, Matthew A., Rugaber, G., Mitra, P.P., Dicholkar, A., "Propagation of Surface-to-LEO Vortex Rings For Orbital Debris Management." In 5th AIAA Atmospheric and Space Environments Conference, p. 2682. 2013.
Key Invited Talks (including upcoming):
SIAM - Learning Operators from Data Symposium 2021, Dallas, TX, 'Non-Intrusive Machine Learning Models for Fluid Simulations'
Parallel CFD International Conference 2020, Nice, France, 'Neural ODEs for Reduced Order Models' - (to be delivered by collaborator, Dr. M. Dias Ribeiro)
AISC Spotlight Talk 2020, 'ML in Climate', [LINK]
GPU Technology Conference, 2020, Santa Clara, CA: 'Climate sub-closures using GPUs' [LINK]
New York Academy of Sciences - Machine Learning Symposium 2020, New York, New York, 'Neural ODEs for Reduced Order Models'
Physics Informed Machine Learning, 2020, Santa Fe, NM: 'Turbulence Forecasting via Neural ODEs'
Google Cloud Platform computing award, 2019. Grant value: $5000
National Science Foundation, Research proposal award 2019, Grant value: $125,000
NeurIPS Travel award 2019, Award value: $1500 + Complimentary registration
Selected to be a citizen-scientist astronaut, Project PoSSUM, class of 2016
Member of the month, UN-Space Generation Advisory Council, 2013
Best Tracking System award, NASA Robotics Mining Competition, 2011