Contact Me
My detailed CV
GitHub Profile
Twitter Profile
Theoritical ML Posts
I am Souradip Chakraborty, currently a 1st year Computer Science PhD student at the University of Maryland, College Park from Fall 2021. My current research focusses on the Theoritical and Practical aspects of Deep Reinforcement Learning especially from a Representation learning viewpoint with a special emphasis on learning State Abstraction. My research also explores Sample-efficient and Safe Model-based reinforcement learning under Uncertainty, noisy-settings and Sparse rewards scenarios. I am also extremely interested in designing Sample-efficient High Dimensional Gaussian Processes and believe that can pave the pathway for Self-Supervised Representation Learning
In the past I have worked as a Research AI Scientist at Walmart Labs, India after completing my Masters from the Indian Statistical Institute in 2018 summa cum laude and Bachelors in Electronics & Electrical Engineering from Jadavpur University, India in 2014. Recognised by Google as a Google Developer Expert in Machine Learning for my open-source contributions & mentorships in the field of AI and ML.
Co-authored several US patents and top-tier publications (COLING, ICPR, ICDM, SIGIR) in the field of AI & ML applications in NLP and Computer Vision domain. Selected as the Youngest Technical Speaker for the very prestigious Data Hack Summit’2019 and 2018 by Analytics Vidhya. Currently,I am also a Thesis Supervisor for students at upGrad’s online Master’s Program in Data Science as well as Machine Learning with Liverpool John Moores University (LJMU)
Mixed Space Representation and Graphical Embedding - We explicitly model the probabilistic dependence structure among the mixed type of variables by an undirected Graph and Spectral decomposition of the graph Laplacian matrix provide the desired feature transformation.
Interpretability of Black-box Models - Worked in the field of finding the most appropriate classifier and classifier conditional hyperparameters given a dataset using Meta-Learning and Gaussian process incorporating the dependencies and relationship among the hyperparameters in the non-convex space and was able to implement it and show explicit relations among the hyperparameters for regularised linear models(Ridge regression) .
Dissertation Thesis - Developed a new methodology for simultaneous monitoring multiple output characteristics using Multivariate regression and Derringer function which got published in Indian Statistical Institute Technical Report 2017-18.
Amrit Singh Bedi, Souradip Chakraborty, Anjaly Parayil, Brian Sadler, Pratap Tokekar, Alec Koppel On the Hidden Biases of Policy Mirror Ascent in Continuous Action Spaces, under submission.
Souradip Chakraborty, Ekaba Bisong, Shweta Bhatt, Thomas Wagner, Riley Elliott and Francesco Mosconi [BioMedBERT: A Pre-trained Biomedical Language Model for QA and IR], accepted at COLING’2020
Saswata Sahoo,Souradip Chakraborty Graph Spectral Feature Learning for Mixed Data of Categorical and Numerical Type, accepted at ICPR’2020
Souradip Chakraborty, Sayak Paul, Aritra Roy Gosthipaty G-SimCLR : Self-Supervised Contrastive Learning with Guided Projection via Pseudo Labelling, accepted at the DLKT Workshop proceedings of IEEE ICDM’2020
Souradip Chakraborty, Ekansh Verma, Saswata Sahoo, Jyotishka Datta FairMixRep : Self-supervised Robust Representation Learning for Heterogeneous Data with Fairness constraints, accepted at the DLC Workshop proceedings of IEEE ICDM’2020
Ekansh Verma,Souradip Chakraborty, Vinodh Kumar Propaganda Fragment Detection using Diversified BERT Architectures based Ensemble Learning, accepted at the SemEval Workshop COLING’2020
Ekansh Verma, Vinodh Kumar, Souradip Chakraborty Deep Multi-level Fusion Learning Framework for Multi-modal Product Classification, accepted at the SIGIR eCom’20 Workshop Proceedings
Ojaswini Chhabra, Souradip Chakraborty Siamese Triple Ranking Convolution Network in Signature Forgery Detection, selected at NCMLAI’19,AICAAM’19.
[US Patent, Walmart Ref.6031US01] - Reverse Engineering Food Ingredient Share estimation using Constrained Optimization by Gregory Dixon, Ojaswini Chhabra, Souradip Chakraborty, Mallikharjuna Maruthi Vallabhajosyula. Provisionally filed, 2019.
[US Patent, Walmart Ref.5928US01] - Retail Based Cost Reverse Engineering and Cost comparison within Item Similarity Clusters for Cost Negotiations by Souradip Chakraborty, Mani Garlapati.
[US Patent, Walmart Ref.5603US01] - System and Method for Detecting Signature Forgeries by Souradip Chakraborty, Ojaswini Chhabra.
[US Patent, Walmart Ref.5118US01] - System and Method For Automated Electronic Catalogue Management and Image Quality Assessment by Souradip Chakraborty, Rajesh Shreedhar Bhat, Mani Garlapati.
[US Patent, Walmart Ref.5008US01] - Generating Customized Alerts with Computer Vision and Machine Learning by Souradip Chakraborty, Rajesh Bhat, Mani Garlapati, Lakshmi Praneetha Kommuru.
[US Patent, Walmart Ref.4970US01] - Architecturally-Distributed Apparatus and Method to Form and Leverage Clustered Content by Souradip Chakraborty,Sunil Potnuru, Mani Garlapati.
Souradip Chakraborty,Amlan Das, Sai Yashwanth Risks and Caution on applying PCA for Supervised Learning Problems published on Towards Data Science,Medium’ 2019
Souradip Chakraborty,Rajesh Shreedhar Bhat Why not Mean Squared Error(MSE) as a loss function for Logistic Regression? published on Towards Data Science,Medium’ 2019.Trending in Machine Learning Category
Souradip Chakraborty Dimensionality Reduction in Supervised Framework and Partial Least Square Regression
Ojaswini Chhabra & Souradip Chakraborty’s Poster on ‘Hierarchical Signature Fraud Detection’ got selected in WiML workshop 2019.
Second Prize for Poster Presentation Judges’ Choice Awards in AI Research Day,IIM Bangalore in association with IBM research Labs’2019.
Lakshmi Praneetha Kommuru, Mani Garlapati, Souradip Chakraborty,Rajesh Bhat’s Customers consumption based Recommendation system, accepted for “Poster Session” at the Grace Hopper Celebration India (GHCI)’18 conference.
[2019] - Technical Keynote speaker for the very prestigious Data Hack Summit’2019 by Analytics Vidhya on ‘Visual Attention and Image Captioning’
[2019] - Thesis Supervisor for students at upGrad’s online Master’s Program in Data Science as well as Machine Learning with Liverpool John Moores University (LJMU) .
[2019] - Key-Note Lecturer to the Faculty of Presidency University,Bangalore at the Faculty Development Program on Statistical Learning Theory and Machine Learning.
[2019] - Key-Note Lecturer to the students of Computer Science department of Coimbatore Institute of Technology at the Machine Learning Workshop on Machine Learning with Python.
[2019] - Technical Keynote speaker for Target Talks AI Session-3 Bangalore’2019.
[2018] - Youngest Technical Keynote speaker for the very prestigious Data Hack Summit’2018 by Analytics Vidhya.
Runners up,Codeception 2019 Walmart Labs International Hackathon - Gradient based constraint optimisation to estimate ingredient’s proportion in foo items.
Rank-13,Crowd Analytix’s Propensity to Fund Mortgages competition 2019 – Implemented LightGBM with error analysis to identify the curvature of the variables and interaction among the features in modelling the response variable.
Bronze Medal,Capillary Machine Learning Hackathon by Analytics Vidhya’2019 – Implemented Alternating Least Squares Method for Implicit recommendation.
Bronze Medal,WNS Analytics Wizard 2018 challenge – An ensemble of Boosting and Deep Neural nets with synthetic minority oversampling was implemented to solve the classification problem with class imbalance.