Pintu Kumar

Ph.D. Candidate in Industrial Engineering & Operations Research

Mumbai, IN.

About

Highly accomplished Ph.D. candidate in Industrial Engineering & Operations Research at IIT Bombay, specializing in advanced Machine Learning and Deep Learning, with a focus on graph data, NLP, and Computer Vision. Proven expertise in developing novel algorithms, leading impactful research projects, and publishing in top-tier conferences, demonstrated by significant contributions to robust node classification and information diffusion modeling. Eager to leverage deep theoretical knowledge and practical implementation skills to drive innovation in complex data science and AI challenges.

Work

Indian Institute of Technology Bombay

|

Teaching Assistant

Mumbai, Maharashtra, India

→

Summary

Supported and instructed graduate-level courses in Machine Learning, Optimization, and Deep Learning, enhancing student understanding and practical application of complex technical concepts.

Highlights

Provided comprehensive academic support for 8 distinct graduate courses including 'Topics in Learning Algorithms' and 'Deep Learning - Theory and Practice', directly contributing to student success and comprehension.

Developed and delivered instructional content for 'Modelling and Computation Lab' and 'Probability and Stochastic Processes II', enhancing students' practical skills and theoretical grounding.

Served as NPTEL PMRF Teaching Assistant for 3 national online courses ('Basic Linear Algebra', 'Essential Mathematics for Machine Learning' (twice), 'Deep Learning'), extending high-quality education to a broad audience.

Facilitated learning for hundreds of students by clarifying complex topics, assisting with assignments, and contributing to the overall academic environment across multiple institutions.

Volunteer

Volunteer

|

IISc Bangalore

→

Contributed to the successful execution of the GAME-ARTS Symposium, supporting event logistics and participant engagement.

Highlights

Assisted with event coordination and participant support at the GAME-ARTS Symposium, ensuring smooth operations for a major academic conference.

Education

IIT Bombay

Mumbai, Maharashtra, India
→
Ph.D. Industrial Engineering & Operations Research Grade: 9.02 CPI

Courses

Introduction to Machine Learning (EE 769)

Mathematical Optimisation Techniques (IE 609)

Foundations of Machine Learning (CS 725)

Deep Learning - Theory and Practice (IE 643)

Topics in Learning Algorithms (IE 683)

Learning with Graphs (CS 768)

Probability and Stochastic Processes II (IE 622)

IEOR Lab (IE 684)

IIT Jodhpur

Jodhpur, Rajasthan, India
→
Post Graduation (M.Sc.) Mathematics Grade: 9.44

Motilal Nehru College, University of Delhi

Delhi, Delhi, India
→
Graduation (Bachelor's) Mathematics Grade: 83.73%

Venkateshwar International School

Delhi, Delhi, India
→
Intermediate (12th Grade) Grade: 92.80%

Venkateshwar International School

Delhi, Delhi, India
→
Matriculation (10th Grade) Grade: 9.8

Awards

Prime Minister Research Fellowship

Government of India

Awarded a prestigious national fellowship supporting doctoral research in India.

IIT Jodhpur Silver Medalist

IIT Jodhpur

Recognized for scoring the highest marks in M.Sc. Mathematics batch 2019.

Qualified GATE in Mathematics

Graduate Aptitude Test in Engineering (IITs/IISc)

Achieved qualifying scores in the Graduate Aptitude Test in Engineering for Mathematics.

Gold Medal in Squash

PGGC (Post Graduate Games Council)

Won a gold medal in Squash while representing the department at the Post Graduate Games Council.

CSIR-UGC(NET) Junior Research Fellowship in Mathematical Science

CSIR-UGC

Secured a competitive national fellowship for pursuing research in Mathematical Science.

Qualified JAM in Mathematics and Mathematical Statistics

Joint Admission Test for M.Sc. (IITs)

Achieved qualifying scores in the Joint Admission Test for M.Sc. in Mathematics and Mathematical Statistics.

CSSS scholarship

Central Sector Scheme of Scholarships

Received a government scholarship supporting undergraduate and postgraduate studies from 2014-2019.

Publications

Impact of label noise on graph homophily

Under Review

This work derives the change in homophily when injected with label noise. Based on this, it proposes a unifying theory for label noise robustness for graphs and examines the impact of different homophily increase techniques on noise robustness.

EDN: A Novel Edge-Dependent Noise Model for Graph Data

ECMLPKDD, Springer Nature Switzerland

Proposed a novel node label noise model where noise is injected via edges and shown to depend on node degree. Designed three EDN variants revealing major performance drops in existing GNNs under edge-dependent noise. Established EDN as a benchmark for graph learning robustness with experimental and statistical validation.

Graph Machine Learning Essentials

Vibrant Publishers, Editorial and Production Phase

Authored a comprehensive book blending theory, algorithms, and practical implementation of Graph Machine Learning using PyTorch Geometric. Covers fundamentals to advanced topics, including node embedding, graph neural networks, graph transformers, practical considerations, and real-world applications.

DeGLIF for Label Noise Robust Node Classification using GNNs

Under Review (Pintu Kumar and Nandyala Hemachandra)

Introduced DeGLIF, a novel denoising framework for node classification on graph data with noisy node labels. DeGLIF uses a small clean set and leave-one-out influence functions for robust node-classification. Developed two variants that identify and relabel noisy nodes without knowing the noise model or noise rate. Provided a theoretical guarantee: detected noisy points provably increase model risk. Validated DeGLIF via extensive experiments showing improved classification accuracy than SOTA algorithms.

Languages

English

Skills

Programming Languages

Python, C++, Matlab, Mathematica, AMPL.

Machine Learning Frameworks

Scikit-learn, PyTorch, PyG, TensorFlow.

Machine Learning & Deep Learning

Machine Learning, Deep Learning, ML/DL for Graph Data, NLP, Computer Vision, ML/DL for real-world data, Graph Neural Networks, Node Embedding, Graph Transformers, Topological Data Analysis, Open Set Recognition, Information Cascade, Node Classification, Link Prediction.

Data Analysis & Modeling

Statistical Validation, Quantitative Methods, Qualitative Methods, Data Collection, Data Analysis, Model Development, Algorithm Design.

Research & Academic Skills

Research, Publication, Theoretical Guarantees, Experimental Validation, Academic Projects, Seminar Presentation, Teaching Assistantship, Course Development.

Interests

Hobbies

Badminton, Numismatics, Philately.

Projects

CGDL for Open Set Recognition (Course Project)

→

Summary

Implemented Conditional Gaussian Distribution Learning (CGDL) for open set image recognition, enabling classification and unknown detection.

Embedding Design During Information Cascade (Course Project)

→

Summary

Implemented Deep Collaborative Embedding (DCE) to model information diffusion on social networks and designed autoencoder-based embeddings.

Topological Data Analysis (Course Project)

→

Summary

Conducted an in-depth study of Topological Data Analysis (TDA) for uncovering structural patterns in high-dimensional data.

Graph Node Embedding (Seminar Project)

→

Summary

Explored core methods for representation learning on graphs, including DeepWalk, Node2Vec, and HARP, for node classification and link prediction.

Dynamics of S-Unimodal Maps (Masters Thesis)

→

Summary

Conducted a comprehensive study of S-unimodal maps to analyze the dynamics of discrete nonlinear systems.

Socio Economic Survey of Students Migrating to Delhi for Undergraduate Studies (Innovation Project)

→

Summary

Conducted a socio-economic survey on students migrating to Delhi for undergraduate studies, funded by the University of Delhi.