Rohit Mujumdar

rohit dot mujumdar at yahoo dot com

I recently graduated from Georgia Institute of Technology with an MS in Computer Science (specialization in Machine Learning). I currently work as a Software Engineer with Business Services Layer organization at NCR Corporation in Atlanta, GA.

I have closely collaborated with Prof. Srijan Kumar in developing a reputation system for Birdwatch, Twitter’s new notes-based, community-driven feature to address misinformation, which was successful in influencing the Birdwatch platform. Our paper based on this work got accepted to ASONAM'21.

I also interned with the Cloud Services Design Analytics group at IBM Research. I filed a patent based on my work and also authored a poster, which won the Best Poster (Honorable Mention) award at INFORMS'20

Prior to commencing my MS degree in Fall 2019, I worked as an NLP Research Engineer at Froot Research. As one of the first few employees of an early-stage AI start-up, I was mentored by the founder Amit Gautam.

I earned a Bachelor's degree in Computer Engineering in 2017 from Vishwakarma Institute of Technology, Pune, where I was advised by Dr. Manasi Patwardhan.

Dancing is one of my hobbies - you can see a few of my dance videos on YouTube. I also like to write about movies and TV shows. You can find a few of my review/appreciation blurbs here. I am an avid practitioner of Yoga and an ardent promoter of mental health awareness.

Feel free to drop me an email or DM me on Twitter; I love making new friends and am always open to interesting conversations!

E-mail  |  Resume  |  LinkedIn  |  Github  |  Twitter

Research Interests

My research interests lie in Applied Machine Learning particularly in Natural Language Processing and Knowledge Graphs. I am especially interested coupling ML/NLP/Graph Learning with social media, online communites and with computational social sciences.

News
Publications
3DSP HawkEye: A Robust Reputation System for Community-based Counter-Misinformation
Rohit Mujumdar, Srijan Kumar
Keywords: social network analysis, misinformation, graph algorithms, adverserial attack
ASONAM 2021

We investigate the robustness of Birdwatch against adversaries and shows that the current Birdwatch system is vulnerable to manipulation attacks. To overcome this vulnerability, we develop HawkEye, a cold-start-aware graphbased recursive algorithm, and show that it is more robust to such attacks.

paper | code | video
3DSP A Heuristic Approach To Compute Service Request Resolution Time (Poster)
Rohit Mujumdar, Pawan Chowdhary, Shubhi Asthana.
Keywords: time series analysis, operations research, ticket resolution, predictive model
INFORMS 2020, Best Poster Award (Honorable Mention Award)

We use statistical analyses and regression-based techniques to predict the resolution times of incident tickets. We employ techniques like dynamic rolling window, auto-regressive window-flip and artificial data creation handle data eccentricities.

poster | video
Selected Projects
3DSP Do Scientific Ideas Originating from more Prestigious Universities Spread Faster?
Rohit Mujumdar, David Kartchner
(Data Science for Epidemiology, Fall 2020. Georgia Tech)
Keywords: epidemiology, microsoft academic graph, natural language processing, epistemiology

  • Investigated the imbalance in the spread of ideas across academic research networks caused due to differences in academic prestige using disease spread models adapted from epidemiology.
  • Assessed if idea spread is driven by connectivity amongst original authors or the explicit prestige of their institution

    • website | report | code | software
    3DSP Can Machines Detect if you’re a Jerk?
    Rohit Mujumdar, Parvathy Sarat, Prathik Kaundinya, Sahith Dambekodi
    (Deep Learning, Fall 2020. Georgia Tech)
    Keywords: natural language processing, deep learning, ai ethics, reddit, computational social science

  • Used language models to assess if we can replicate the sentiments shared by Redditors and classify the Redditor's original post according to the verdict that was declared by rest of the Redditors.
  • Attempt to understand how a machine performs in a task that is entirely subjective but is possibly objective

    • report | code
    3DSP Conference Paper Acceptance Prediction
    Rohit Mujumdar, Rohan Goel, Arthita Ghosh, Shravani Sistla, Neha Pande.
    (Machine Learning, Spring 2020. Georgia Tech)
    Keywords: natural language processing, machine learning, feature engineering, peer read dataset

  • Investigated the role of peripheral features of research papers in their potential acceptability
  • Devised several innovative features such as abstract novelty and complexity, research strength score, title word-cloud etc

    • report | code | video
    3DSP Semantic Search Engine using a Dynamic Ontology
    Rohit Mujumdar, Poshraj Sharma, Pranjal Patil, Akanksha Patil, Dr Manasi Patwardhan.
    Keywords: natural language processing, semantic search, ontology, triplet extraction, entity-relation-entity, phrase2vec, education technology
    (Undergrad Capstone Project, 2016-17, VIT Pune)

  • Developed an e-learning platform for government schools by implementing a dynamic science ontology to store triplets (Entity-Relation-Entity) extracted from web-scraped data.
  • Devised a Phrase2Vec model driven similarity-scoring algorithm to replace similar Relations by a representative Relation.

    • report | video | code
    Leadership

    In The Media
    Memberships

    "In the midst of winter, I found there was, within me, an invincible summer." ~ Albert Camus
    Design inspired from here

    Flag Counter