About Me

I am a final year student pursuing a Bachelor's degree in Computer Science Engineering at SRM Institute of Science and Technology in Chennai, India.

My research interests revolve around deep learning architectures, generative models and their architectural complexity, human brain & connectomics, and mechanistic interpretability for AI safety. I'm particularly interested in applying computational approaches to understanding neural systems and contributing to brain-inspired AI development.

I am currently working as an R&D Intern at SPAN Inspection Systems' R&D department on Advancing Coronary artery segmentation and Blockage detection using Computer Vision.

Publications & Ongoing Research 🔬

研究

ResformerAF: Integrating Deep Learning Models for Atrial Fibrillation Detection Using ECG

Under Review

Spearheaded a comparative analysis of ResNet, ResNet with attention, Bi-LSTM, and Bi-LSTM with attention for ECG classification. Proposed a novel ResNet + encoder-based architecture to enhance feature extraction and classification accuracy.

Coronary Artery Segmentation, Blockage Detection and Measurement

In Progress

Leading the development of a deep learning system for coronary artery segmentation and stenosis detection in medical images. Leveraging state-of-the-art models to improve diagnostic accuracy for Coronary Artery Disease (CAD).

Cross-lingual Semantic Equivalence in Large Language Models

In Progress

Investigating how Large Language Models (LLMs) process semantically equivalent prompts in multilingual and code-switched contexts (e.g., Hinglish). Examining semantic preservation across languages to better understand LLM reasoning and improve cross-lingual capabilities.

Projects

作品

AI Startups & Investors Network Visualization

Developed an interactive network graph visualizing complex relationships between AI startups and their investors. Implemented timeline animation features to demonstrate the evolution of the AI investment landscape over time. Created interactive controls for filtering, zooming, and accessing detailed information.

View Project

Leaf Disease Detection using Fine-Tuned SAM2_tiny

Fine-tuned Segment Anything 2 (SAM2_tiny) to enhance automatic segmentation of diseased leaf regions. Built a deep learning pipeline that improves semantic segmentation accuracy and robustness across plant species. Optimized model inference for real-time disease detection applications in precision agriculture.

GitHub

Credit Card Approval Prediction

Engineered an advanced stacked ensemble model combining Gradient Boosting, Random Forest, AdaBoost, and Neural Networks for high-accuracy credit approval predictions. Applied feature engineering techniques like one-hot encoding, scaling, and PCA to optimize model performance. Evaluated model effectiveness using ROC-AUC, precision-recall curves, and confusion matrices.

GitHub

Research Interests

興味

Deep Learning Architectures

Generative Models and Architectural Complexity

Human Brain & Connectomics

Mechanistic Interpretability and AI Safety

Deep Learning Architectures

My foundation lies in developing and understanding deep learning models, evidenced by my work with ResNet, LSTM, Bi-LSTM, and UNet architectures. This experience has provided me with a strong understanding of complex neural networks and their applications in areas like image recognition and sequence modeling.

Generative Models and Architectural Complexity

I am particularly fascinated by the architectural innovations in generative models like Stable Diffusion, especially their use of VAEs and UNet. These models represent a significant leap in architectural complexity, and I am eager to contribute to their further development and understanding.

Human Brain & Connectomics 🧠

My interest naturally extends to the most complex system we know – THE HUMAN BRAIN. Inspired by advancements in connectomics, I am captivated by the brain's design and architecture. I believe that insights from neuroscience and connectomics can offer invaluable inspiration for the next generation of AI architectures.

Mechanistic Interpretability and AI Safety

I believe that truly understanding how complex systems arrive at their outputs is crucial for both advancing the field and ensuring AI safety. My current independent research into how LLMs process semantically equivalent prompts across languages directly reflects this commitment to interpretability.

Motivation

動機

Research, like bonsai cultivation, requires patience, vision, and meticulous attention to detail.

Kung Fu Panda Meditating Under Cherry Blossom Tree

"Your mind is like this water. When it becomes agitated, it becomes difficult to see. But if you allow it to settle, the answer becomes clear."

- Master Oogway