Abhishek K Dubey

Postdoc, Oak Ridge National Lab, Biomedical and AI initiative

Ph.D., Computer Science, Duke University

B.Tech., Computer Science and Engineering, IIT Kharagpur

Abhishek

Research Summary

My research focus is on developing efficient and accurate algorithms for data analysis in biomedical research. I wish to understand the foundations of data analysis and idiosyncrasies of biomedical data towards developing scientifically interpretable algorithms. Originally trained as a computer scientist, I have studied data analysis in biomedical research for over 6 years during my Ph.D. and postdoctoral research. My research interest is to develop algorithms for biomedical applications such as medical imaging, bioimaging, and public health research. My previous and current work has involved several major areas: (a) Developing novel inversion algorithm for deformation field with theoretical guarantee of accuracy and convergence; (b) Developing alternating symmetric completion algorithm for image registration; (c) Designing statistical features of time-series data extracted from single-molecule fluorescence images for barcoding; (d) Developing probabilistic deep learning model for information extraction from medical reports; e) Model reduction of shallow CNN model for reliable and interpretable deployment; (f) Developing domain detection and adaptation algorithm for generalization of DL models to multi-center dataset. I have authored 10 scientific papers and 5 abstracts in peer-reviewed conferences and journals. A complete list of my previous publications is included in my full CV. I am also serving as a reviewer in scientific journals Journal of Biomedical and Health Informatics , Neural Processing Letters, and IEEE access.

I am on the job market (2020-21). You can find my research statement here and slides here.

Research Interests

  • Deep learning: model reduction, domain adaptation, information extraction, distributed training

  • Image analysis: symmetric registration completion, inversion of deformation field via bi-residual iteration, domain adaptation method for diagnosis of lung conditions from multi-center chest x-ray dataset

  • Text analysis: deep kernel learning for information extraction from medical reports, model reduction of information extraction model for its reliable deployment

Profile Highlights

  • Postdoctoral researcher in AI, ORNL, current

  • Ph.D. in Computer Science, Duke University, 2012-2018

  • Research intern at IBM Research, Almaden, summer 2014

  • Oracle India, Technical staff member, 2009-2012

  • B.Tech in Computer Science, IIT Kharagpur, 2005-2009

  • Research intern at EPFL Switzerland, summer 2008

Contact

Office: D104, 5700 building, 1 Bethel Valley Rd, Oak Ridge, TN 37830