Deep learning project using a pre-trained CNN in Caffe to colorize images by treating it as a classification task in the LAB color space, improving accuracy over traditional methods.

A neural network implemented from scratch using NumPy without any ML frameworks, exploring the underlying math and matrix calculus, and applied to classify digits from the MNIST dataset.

A deep learning project that detects defects in transmission line insulators using YOLOv8 (Ultralytics). It uses a Roboflow dataset and runs on Google Colab for efficiency.

Mathematical Analysis of Neural Networks for MNIST

Implementing MNIST Classification with NumPy and Matrix Calculus

Detecting Transmission Line Faults with YOLOv8

Our electric grid is outdated, and not enough people are talking about it

Deep Learning Techniques for Image Colorization: A Step-by-Step Guide

Colorizing Grayscale Images Using Caffe and Deep Learning Models

Diffusion Models: How AI Turns Noise Into Art

Exploring Denoising Diffusion Probabilistic Models (DDPMs)

Saving $300M+ Globally With Optimized Model Deployment | Microsoft x TKS

Developed a proposal for an acceleration framework that optimizes ONNX model deployment on hardware, reducing inference costs by up to 40%, and potentially saving $300M annually.

The $2.3 Billion Cost of IKEA's Communication Breakdown | IKEA x TKS

Developed a proposal for an internal communications software, Axonify and FIKA, projected to increase global revenue by $13.5M at IKEA.

Reducing Medical Errors with Programmable Materials in MedTech

Pulse oximeters are 3x more likely to give falsely high readings for people with darker skin, contributing to healthcare disparities. Plasmonic materials can improve accuracy by its property of scattering light to penetrate the skin. Best Overall @TKS Toronto Hackathon

Data Storage Solution - DNA

Data storage costs have surged by 205% over the past five years. DNA storage is a promising energy storage solution, as it is less expensive, more energy-efficient and longer lasting. Top 5 @TKS Global Hackathon