Projects

AdaComp - Adaptive Multi-Tier Compression Middleware for MPI Collectives

MPI middleware with adaptive compression and pipelined communication overlap, reducing network overhead and accelerating collective operations in HPC clusters.

C/C++, MPI, HPC

Knowledge-Aware Graph Prompt Learning for Cross-Domain Recommendation

Cross-domain recommender using PyTorch/DGL with TransE and GAT, introducing adaptive prompt scaling and scalable training over 923K+ nodes.

PyTorch, DGL, GAT, TransE

Real-Time Attack Detection for Networked Systems Under Cyber Threats

Cyber-attack detection and synchronization control system for dynamical networks using ODEs, graph theory, and LMI-based optimization.

ODEs, Graph Theory, LMI Optimization

Skin Lesion Classification using Deep Learning with Attention Mechanisms

CNN-based skin lesion classifier on the HAM10000 dataset using CBAM attention and Focal Loss, achieving robust multi-class classification across 7 disease types.

CNN, CBAM, Focal Loss, PyTorch

Semantic Video Shot Detection and Keyframe Extraction using CLIP

CLIP-based video understanding pipeline for shot detection and keyframe extraction using embedding similarity and clustering methods.

CLIP, Embeddings, Clustering