Bio Summary
AI Research Scientist and Data Scientist with over 4 years of experience in applied probability, statistics, and machine learning. Specialized in Bayesian neural networks, time series modeling, and industrial research. Enthusiastic about AI and continuous learning. Proven track record in developing innovative ML models, leading projects, and collaborating with R&D teams.
Education
Polytechnique Montréal | Sep 2018-Dec 2022
- Ph.D. in Civil Engineering: Machine Learning Specialization – GPA 4/4.3
- Research Grant: Hydro Québec, NSERC
McGill University | Sep 2015-Mar 2018
- M. Eng. in Civil Engineering: Structural Engineering – GPA 3.88/4
- Research Grant: Graduate Research Assistantship
Assam Engineering College | Aug 2010-June 2014
- B. Eng. in Civil Engineering: First-Class Honours with 75.78%
- Research Grant: North-Eastern Council Merit Scholarship
Research Focus
Advanced Machine Learning Research
- Bayesian Neural Networks, Regression, Classification, Time Series
Industrial Research Leadership
- End-to-end ML solutions, process automation, software training
Machine Learning Research Scholarship
- Neurocomputing, Adaptive Control, Signal Processing
AI Enthusiast
- Diverse end-to-end deep learning projects in CV, NLP, Generative AI
Professional Journey
Machine Learning Researcher | Polytechnique Montréal | Sep 2018 – Present
- Developed AGVI for uncertainty quantification in Bayesian neural networks.
- Improved training time significantly, applied in regression, classification, and time series.
- Published AGVI in the International Journal of Adaptive Control and Signal Processing.
Research Consultant | Polytechnique Montréal | Sep 2018 – Present
- Developed ML approach for nonlinear dependencies in time series.
- Collaborated with dam engineers for real-time anomaly detection, winning competitions.
Honors & Awards
Eliminated hyper-parameter tuning and sped up Bayesian neural network training.
First and fourth rankings in predictive modeling competition at ICOLD-BW2022.
Author Metrics
Number of publications: 4
Conferences and Talks: 2
First-author publications: 2
Research Timeline
2018-2022: Ph.D. in Civil Engineering with a focus on Machine Learning.
2018-Present: Machine Learning Researcher at Polytechnique Montréal.
2018-Present: Research Consultant for industrial projects in collaboration with Hydro Québec.
2022: Published AGVI in the International Journal of Adaptive Control and Signal Processing.
2022: First and fourth rankings in predictive modeling competition at ICOLD-BW2022.
Publications & Contributions
Journals:
- Tractable Uncertainty Quantification in Bayesian Neural Networks (Neurocomputing, 2023).
- Gaussian Variance Inference for State-Space Models (IJACSP, 2023).
- Inspector’s Uncertainty Inference Using Network-Scale Visual Inspections (J. Computing in Civil Engineering, 2023).
- Gaussian Multiplicative Approximation for State-Space Models (Structural Control and Health Monitoring, 2022).
Conferences and Talks:
- Online aleatory uncertainty quantification for probabilistic time series (ICASP14, 2023).
- Dam Behavior Prediction Using Bayesian Models (Benchmark Workshop on Numerical Analysis of Dams, 2022).
Publications Top Noted
1. Building Classification Scheme and Vulnerability Model for the City of Guwahati, Assam
- Authors: J Pathak, R Bharali, B Deka, S Pathak, IJ Ahmed, DH Lang, A Meslem
- Published in: EQRisk Project Report
- Year: 2015
- Cited By: 6
2. The Gaussian Multiplicative Approximation for State-Space Models
- Authors: B Deka, L Ha Nguyen, S Amiri, JA Goulet
- Published in: Structural Control and Health Monitoring
- Volume: 29 (3)
- Page: e2904
- Year: 2022
- Cited By: 5
3. Damage Assessment of RC Frame Structures under Long Duration Aftershock Ground Motion
- Authors: B Deka, SN Rahman, P Tamuly
- Published in: International Journal of Innovative Research in Science, Engineering, and Technology
- Volume: 3 (9)
- Pages: 16144-16149
- Year: 2014
- Cited By: 5
4. Analytical Bayesian Parameter Inference for Probabilistic Models with Engineering Applications
- Author: B Deka
- Published at: Polytechnique Montréal
- Year: 2022
- Cited By: 3
5. Analytically Tractable Heteroscedastic Uncertainty Quantification in Bayesian Neural Networks for Regression Tasks
- Authors: B Deka, LH Nguyen, JA Goulet
- Published in: Neurocomputing
- Pages: 127183
