HoleyMoley: Senior Design Capstone Project (Video Game)
January - May 2022 [Game Information and Download]

  • Designed a multiplayer, 2D side-scrolling platformer game as a senior design capstone project using XNA framework in C#.
  • Built the physics engine, lighting engine, and rendering engine.
  • Designed the core game mechanics and helped building networking architecture.

Multi-Seed Consensus Reference-Based Mapper (MSCRM)
September - December 2021 [Source] [Presentation]

  • A reference-based read-mapper which performs ungapped alignment of paired sample reads with single nucleotide variants (SNVs) and spike-in contaminants on reference sequence.
  • Results of mapping reads from SARS-COV-2 sequence:
    • Mapped 6,657,204 reads in 20.52 seconds, and thus, achieved 19,465,509 reads/min throughput on a 12-core/24-thread Ryzen 9 5900X CPU machine.
  • Achieved recall of 0.999 and precision of 1.000.

Predicting Satisfaction in Live Customer Support Chats Through NLP
January - May 2021 [Project Video]

  • Placed second in Rice 2021 D2K Showcase
  • Developed a real-time, deep-learning/logistic regression hybrid model to monitor customer service chats and dynamically predict the probability of a positive outcome.
  • Successfully flagged 75% of negative outcomes at the midway point of an ongoing conversation.

Rice Eclipse Software for Engine Testing
August 2019 - May 2021 [Source] [Project Video]

  • Won “Best Aerospace or Transportation Technology” award in Rice 2021 Engineering Design Showcase
  • Wrote a hybrid rocket engine control software in C++ that actuates valves, collects raw sensor data, performs ignition, and communicates with a remote dashboard using UDP and TCP protocols with average 0.2% packet loss.
  • Designed a mission control dashboard in JavaScript using Node.JS and Sockets.IO that calibrates, normalizes, and plots the incoming raw data, and the analysis software performs high-pass filtering using Pandas on Python.

Using Facial Recognition Methods for Large-Scale Multi-Class Landmark Classification Problem
August - December 2020 [Source] [Presentation]

  • Built a convolutional neural network with EfficientNet-B7 feature extractor for multi-class landmark classification. The model is trained using ArcFace and Mini-Batch SGD.
  • Achieved a 0.998 micro Average Precision score when dataset is limited to 2 labels (binary classification).
  • Achieved 0.579 micro Average Precision score when the entire dataset is used (2,015 unique labels).

Traffic Analytics Software
February 2020 at TAMUhack 2020 [Source]

  • Placed 1st out of 20 entries in StateFarm challenge for the best data science application.
  • Developed an interactive map to analyze traffic accidents and their correlation with road features in the U.S. on Python.

Virtual Banking Assistant
September 2019 at HackRice 9 [Source]

  • Placed 1st in Bill.com challenge for the best API integration and won MLH award for the best use of Google Cloud.
  • Built a virtual banking assistant that works with speech recognition using DialogFlow and TensorFlow.
  • Prepared and provided corpus to train HMM based machine-learning algorithm for labelling.

Food Desert Map
September 2018 at HackRice 8 [Source]

  • Won Indeed Challenge for developing technologies that aims to solve societal problems.
  • Built a search engine that shows whether entered location is within a food desert using Google Maps and Search API.