This report focuses mainly on the architecture of Transformers, motivated by my fine-tuning project.
Limitations: The report has not included multimodal features and reasoning abilities using reinforcement learning.
This is a tutorial providing a step-by-step guide on running light-weight LLMs locally on Android using Termux. It also explains the limitations with iOS support.
A detailed report on fine-tuning the DeepSeek-R1 model on a reasoning dataset, including performance benchmarks and insights into model improvements.
The fine-tuning and benchmarking methods are posted on GitHub. Slides keeping track of the project timeline and results are also attached.
An implementation of a convolutional neural network that classifies various animals. The project covers dataset preparation, model architecture, and performance evaluation.
This project explores real-time object detection using OpenCV. It includes details on setup, algorithm selection (Haas Cascades and its variations are being used).
The GitHub repository also includes some basic operations and manipulations on images and videos.
An experimental project focused on developing an autonomous car system, where I collaborated with 3 other teammates. The project demonstrates sensor integration, in which we are given only 2 sensors, and autonomous decision-making.
A work in progress project exploring the integration of multimodal features into large language models, aiming to enhance user interaction with both text and visual data, especially images.