Don't worry, I won't eat you.
My name is Prannaya "Prannay" Gupta and I am the fastest smartest weirdest human alive. To the outside world, I'm an ordinary NUS High student, but secretly, with no help from my friends at NUS High, I write code and publish findings regularly. When I was a child, I started coding something(s) impossible. I spent my hours coding away. Then the stress of the school caused me to do something impossible. I brought new problems to myself, and I am the only person stupid enough to stop them. I am ThePyProgrammer.
But seriously, call me Prannay.
Yes, this is turning into a resume. I totally don't have one of those.
Repository |
Information |
Still Active? |
|
My first research project, where we used signal processing techniques with numerical integration and Fast-Fourier Transform methods utilised and Support Vector Machines of varied configurations on various variables of tri-axial accelerometer inertial measurement unit (IMU) data to identify freezing of gait, a debilitating condition amongst Parkinson's Patients. We also developed a semi-functional wearable prototype with an Arduino Nano BLE and an Android Application capable of keeping track of movement and interfacing with Firebase and online servers to analyse this freeze data. Done as part of the MOE Science Mentorship Programme at NUS. This project clinched us a Gold Award for the Singapore Science and Engineering Fair. |
Yes. |
|
Utilised PyTorch and standard image preprocessing techniques, with transfer learning on the pre-existing ResNet-UperNet encoder-decoder architecture trained on CSAILVision's MIT ADE20K dataset, to create two models on par with industry standards for semantically segmenting images of HDB apartment interiors and exteriors into various appropriate classes to gain a greater understanding of the structure and layout of various homes in Singapore. Done as part of the Research Mentorship Programme at SUTD. |
No. |
|
CS5131 Project to identify and analyse various writing components in textual pieces, such as argumentative essays and opinion editorials, in order to draft a system to provide feedback to students. Utilised PyTorch and HuggingFace Models with transfer learning on the small labelled sample of argumentative essays from Kaggle as well as the noisy student approach on unlabelled opinion editorials from The New York Times amongst others. Developed a Vue.js and Flask application to interface with these models, with the option to add text and segment it into various elements via a data table. |
No. |
|
Utilised multiple methods to perform simulations (via Runge Kutta) and photometric classifications (via complex signal processing algorithms) to identify Three-Body systems via light curves and luminosity data from the pre-existing Kepler and K2 Missions. Although unsuccessful, this project gave a preview of the atrocities of astronomical research. |
Yes. |
|
Utilised Tensorflow models and OpenCV haarcascades to create a prototype via Tkinter (for desktop) and Flask (for web) that analyses and identifies faces, and then processes them based on emotion. This project clinched us a Judges' Choice Award at the BuildingBloCS 2021 June Conference. |
No. |
|
CS4232 Project to Analyse Light Pollution Data and Find Patterns in Country-Based and State-Based Light Pollution Aggregates, including geospatial analyses with .tif files from the NASA DMSP OLS Dataset openly available. |
No. |
PS: If you didn't realise, the first paragraph is just the description of the starting of any "The Flash" episode.
Updated: Back when it actually mattered.