Fahmid Morshed Fahid

I am a Ph.D. student in the IntelliMedia Lab at North Carolina State University , under the supervision of Dr. James Lester . My research interest is in Intelligent Learning Environment domain to understand pedagogical learning using Reinforcement Learning, Educational Data Mining and Mechine Learning.

Before joining NC State, I worked as a Software Engineer in Research and Development department of Reve System Ltd. . I obtained my Bachelors Degree in Computer Science and Engineering from Bangladesh University of Engineering and Technology (BUET) .

Email  /  Alternate Email  /  Resume  /  GitHub  /  LinkedIn  /  Twitter


My research interest is to understand Pedagogical Learning in Intelligent Learning Environemnt using Reinforcement Learning, Educational Data Mining and Machine Learning techniques to produce real-world optimized policies. I believe that the next decade will find better, cheaper and simpler ways to process and utilize data-driven implementation of Reinforcement Learning (specifically, offline learning) and will find an optimized human-in-the-loop framework for understandably, controllability and reliability of deep networks.

I am currently working with off-policy Deep Reinforcement Learning on Tutorial Planning Agent using real world Amazon Mechanical Turk data. This project studies ICAP model of cognitive psychology to understand learning gain. The goal is to produce effective remediation policies using only historical data (without an environment to act).

  • Better Technical Debt Detection via SURVEYing, 2019, Pre-print
  • TERMINATOR: Better Automated UI Test Case Prioritization (collaboration with LexisNexis), 2019, ESEC/FSE 2019
  • Software Engineering for Fairness: A Case Study with Hyperparameter Optimizationn, 2019, Pre-print
  • Industrial Experience

    Software Engineering Intern at Power Costs Inc.

    May 2019 - August 2019 (Raleigh, NC)

    Worked as a member of Project Engineering team. Developed six SVN repository organizing and monitoring tools for PE team. Created an SVN Splitter that splits projects into new repositories and wipe-out bad commit histories. Saved approximately 10% storage in AWS.


    Software Engineer at Reve Systems Ltd.

    March 2017 - August 2018 (Dhaka, Bangladesh)

    Full Stack Agile development using JDBC and JSP framework. Maintained Linux server, MySQL database, Packet Scheduling for 20 commercial projects. Developed one e-commerce websites, one digital evedance management system for Bangladesh Government, one CRM software.

    Teaching Assistantship Experience

    CSC510 - Fall 2019 (Software Engineering) at NCSU


    CSC216 - Fall 2018 (Programming Concepts - Java) at NCSU

    prl image collected from wiki

    Deep Reinforcement Learning on ICAP Based Tutorial Planning

    Research Project (NCSU) - Deep Reinforcement Learning, Deep Q Network, Offilne Learing, Intell Coach-RL, Data Mining, Machine Learning

    Building off-policy adaptive remediation agent using deep Q networks to compare against ICAP model from cognitive psychology. Solving credit assignment problem and sparse reward problem in a MTurk data driven offline learning environment.

    prl image collected from wiki

    Active Learning on Source Code Comments

    Research Project - Active Learning, Transfer Learning, Data Mining, Machine Learning, Hyper-parameter Optimization, Multi-processing, Classification, Labeling

    Used active learning and transfer learning on total recall problem to analyze the reliability of ground truth labels in source code comments for Self Admitted Technical Debt. Finds 83% recall by utilizing only 16% data.


    Large-scale NLP Analysis on StackOverflow

    Research Project - Data Mining, Machine Learning, Word Embedding, Hyper-parameter Optimization, Multi-processing, Clustering, Classification

    Developed clustering based SVMmodels with automated hyper-parameter optimization to understand the similarities between posts using word embedding. 10% speedup and 5% better F1 score over the baselien of over the baseline of Majumdar et al. 2018.


    NLP Analysis to predict the impact of TedTalk

    Course Project (CSC522 Automated Learning and Data Analysis) - Data Mining, Machine Learning, Word Embedding, Hyper-parameter Optimization, Noise Reduction, Feature Reduction, Classification

    Analyzed different regression and classification approaches to predict the impact of TedTalk using heterogeneous data merging. Implemented the complete data-mining pipeline such as feature reduction, normalization, vectorization, noise reduction etc.


    Visualisation Library Comparision: Matplotlib vs Plotly

    Course Project (CSC510 Software Engineering) - Static Analysis, Code Smells, Repository Mining, Statistics, Python

    Compared the Code Smells, Runtime, Render-time, Community and Repository structures etc. Findings: although Plotly offeres many state-of-the-art functionalities like cross language integration, web integration, interactive graphs etc., Matplotlib is still a Safer (Code Smells and Oper-source community) and Cheaper (Runtime and Render Time) choice over Plotly.


    Web Project Environemnt Hub

    National Hackathon Project (2016) - PHP, MySQL, HTML, CSS, Codeigniter, Bootstrap

    Two day National Level Hackathon in Bangladesh. This website was designed to show the trash location, current environemtal hazards. Two different types of users were used. One Admin type, who manages the databases and updates, while a regular user, who can post new hazard locations. This project was in the Top 10 Projects under Environemnt Category.


    Java Game From Scratch Jumping-Ju

    Course Project (BUET) - Java, OOP, Game Engine, Animation, Sprites, File Handeling

    A course project on OOP. A platform game created from scratch. Learned about game loop, animation, game physics etc.

    Website Template Credits Last updated: 01/02/2019