CHIRADEEP ROY

**** ******** **** *** ***, Dallas TX 75252 646-***-**** adjnxz@r.postjobfree.com

Google Scholar LinkedIn

EDUCATION

Doctor of Philosophy (PhD.) in Computer Science 2016 - present The University of Texas at Dallas GPA: 4.0/4.0

B.Tech. in Computer Science and Engineering 2009 - 2013 VIT University, Vellore (India) GPA: 8.9/10.0

SKILLS

Programming Experience: C, C++, Python, Matlab, R, Bash Operating Systems: Windows, OSX, Linux

Databases: MySQL, Oracle

Libraries/Tools: NumPy, SciPy, Scikit-Learn, Matplotlib, Keras, Weka, CoreNLP Research Areas: Graphical Models, Tractable Models, Time Series, Bayesian, XAI EXPERIENCE

Graduate Research Assistant, University of Texas Jan 2018 - present at Dallas

• Conducted independent research on probabilistic machine learning problems

• Areas: Probabilistic Graphical Models, Tractable Models, Explainable AI (XAI) Risk (Data Science) Intern, Elevate Credit Jun - Aug 2019

• Devised and compared metrics to measure the consistency of XGB Explainer

• Devised a model-agnostic method for scoring and ranking features using single counterfactuals

• Modeled fraud data as a Gaussian mixture model which was used to efficiently compute Bayesian probabilities over features as explanations

• Used: Python, numpy, scikit-learn, matplotlib, R, ggplot2 Graduate Teaching Assistant, University of Texas Aug 2016 - Dec 2017 at Dallas

• Graded assignments, conducted office hours and prepared homework solutions for selected courses

• Courses: Data Structures & Algorithms, Database Systems, Artificial Intelligence Programmer Analyst, Cognizant Technology Solutions Aug 2013 - Aug 2016

• Remediated SAS programs on Windows/Linux

• Used: SAS, VBScript, bash

RESEARCH PROJECTS

Rule-Extraction for Multi-Label Tractable Models Sep 2020 - present

• Developed a generic and explainable framework for multi-label classification problems

• Developed an algorithm to extract soft probabilistic rules from the learned model

• Devised metrics for measuring goodness of rules against baselines such as TREPAN Dynamic Cutset Networks Sep 2019 - Sep 2020

• Developed a novel, expressive and tractable framework for modeling sequential/ temporal data based on cutset networks

• Formally proved tractability guarantees for certain constrained models and semi- tractability for unconstrained models

• Empirically verified the efficacy of our model w.r.t. state-of-the-art models such as DBNs, LSTMs and DSPNs

• Used: Python, numpy, scikit-learn, keras

Explainable Activity Recognition in Videos Jun 2018 - Mar 2019

• Created a video activity recognition system using a two-layered architecture

• Bottom layer used a deep CNN based on GoogleNet to assign noisy activity labels to each video frame

• Top layer used a probabilistic model to both correct these noisy labels, model temporal dependencies as well as add a layer of explainability

• Used: Python, numpy, scikit-learn

ACADEMIC PROJECTS

Structure Learning in Markov Networks using Forced Pruning May 2018 Explored all Markov Networks having a fixed model complexity M and optimized heuristically using greedy selection and rejection sampling. (C++) Weighted Graph Coloring using EM Mar 2018

Defined a weighted distribution on the edge coloring problem in graphs and used Expectation Maximization to compute the probabilities for unobserved edges. (Matlab) News Article Search Index Dec 2017

Built a search index over a corpus of over 1,000 articles having over 10,000 words. Used weighted NLP features such as lemmas, POS tags, hypernyms, meronyms, etc. to improve performance. (Java, Maven, SolR, Stanford CoreNLP, MIT Wordnet) Netflix Recommendations using Collaborative Filtering Mar 2017 Implemented a collaborative filtering algorithm to predict user preferences and ran it on the Netflix Prize Dataset. (Python, numpy)

PUBLICATIONS

1. Roy, C., Rahman, T., Dong, H., Gogate, V. and Ruozzi, N. Dynamic Cutset Networks. (accepted at AISTATS 2021).

2. Nourani, M., Honeycutt, D.R., Block, J.E., Roy, C., Rahman, T., Ragan, E.D. and Gogate, V., 2020, April. Investigating the Importance of First Impressions and Explainable AI with Interactive Video Analysis. In Extended Abstracts of the 2020 CHI Conference on Human Factors in Computing Systems (pp. 1-8). 3. Roy, C., Shanbhag, M., Nourani, M., Rahman, T., Kabir, S., Gogate, V., Ruozzi, N. and Ragan, E.D., 2019. Explainable Activity Recognition in Videos. In IUI Workshops.

Contact this candidate