+1 774 208 6637

©2018 by Suvodeep Majumder

HELLO! I’M SUVODEEP MAJUMDER

 

BIO

About Me

I am a PhD Student and a data scientist, with a research focus in the application of AI techniques in Software Engineering at North Carolina State University. I work under Dr. Tim Menzies as part of RAISE Lab.My research explores ways in which aspects of pattern recognition and artificial intelligence can be used to generate Actionable Analytics using Big Data from Different Software Projects.


More specifically, my work focuses on transfer learning by creating a generalized model(aka "Bellwether") to identify a group of exemplary source projects to generalize conclusion drawn from them and apply those learnings in other projects. I also work AI fairness by finding ways to identify group bias in machine learning models and use multi-goal optimization techniques to mitigate them.

In my free time I like to explore the culinary world. 

 

PUBLICATION

Whats already done!!

 

500 TIMES FASTER THAN DEEP LEARNING

MSR, 2018

In this paper we tried to explore alternative method than deep learning for a text mining problem using StackOverFlow dataset, by clustering the data first then tuning machine learning models on each cluster to create local models, which produces a model almost 500 - 1000 times faster with similar performance score.

RESEARCH

What I am working on

 

SOCIO-TECHNICAL GRAPH MINING

Working on a system to create code interaction graph and social interaction graph to understand different aspects of Open Source Projects

2016.png
Screen Shot 2018-11-18 at 9.20.11 PM.png

AUTOMATED LEARNING

This project is to create an automated tool, that incorporates different types of machine learning models, feature selectors with hyper-parameter optimizer to create models for text dataset and return best possible outcome by evaluating the dataset.

ACTIVE LEARNING

This research was to create an active learning model, to learn from very small sample of data to begin with, in a domain where it is very expensive to collect the data labels. Then to understand the attributes which are essential for the classification and as new data is collected, the model only asks for labels which it think is interesting or non interesting depending on model setup.

Screen Shot 2018-11-18 at 9.43.22 PM.png
BUBBLE.png

TRANSFER LEARNING

Developing a new hierarchical transfer learning method using hierarchical clustering and bellwether method to identify exemplary projects in a community to get generalized actionable conclusions to mitigate conclusion instability.

AI FAIRNESS

Working on identifying and mitigating group bias introduced by machine learning models on protected attributes using multi-goal optimization techniques.

1_cc8OWxqKFXje4d_1eYrQkg_edited.jpg

MY EXPERIENCE

DATA SCIENTIST INTERN, IBM

June 2018 - August 2018

As part of IBM DevOps insight team, I was responsible for handling a large set of Open Source and Enterprise project data and try to create actionable analytics, that can be utilized in the product by generating different machine learning models to help development teams work more efficiently.

TEST ANALYST, INFOSYS LIMITED

March  2013 - June 2017

As part Infosys Independent validation System team, I was responsible for managing and working on an agile team which was  responsible for managing web and mobile application testing(system, Automation, Performance).

 

EDUCATION

NORTH CAROLINA STATE UNIVERSITY

August 2019 - May 2023

I am completeing my PhD in Computer Science with research fouce in applicatio of AI in Software Engineering.

NORTH CAROLINA STATE UNIVERSITY

August 2017 - May 2019

I completed my master degree from North Carolina State University with a specialization in Software Engineering and Machine Learning.

WEST BENGAL UNIVERSITY OF TECHNOLOGY

August 2008 - May 2012

I Completed my under graduation from WBUT in 2012 with a Major in Computer Science & Engineering.

 
Man with Briefcase

"If you want something you've never had, you must be willing to do something you've never done"

Thomas Jefferson

 
Clean Space

LET’S CONNECT