Avi Lochab

Motivated and principled Master’s student in Computer Science with a strong foundation in pattern recognition, NLP, and computer vision. Respected for leading with quiet confidence, moral clarity, and a deep commitment to inspiring those around me. Known for a disciplined work ethic, administrative acumen, and a heartfelt drive to uplift through intellect, empathy, and unwavering resolve.

Decentralized Federated Learning Platform – Winner, SpartaHack 9

• Built a decentralized FL platform using PyTorch for collaborative training across 50+ clients

• Enabled private model sharing across financial datasets; achieved rapid convergence with minimal validation loss

Multimodal Brain Tumor Segmentation with DCGANs

• Developed a conditional GAN based segmentation model for brain tumor classification using MRI slices, integrating synthetic data augmentation to address class imbalance

• Achieved significant performance gains by combining adversarial training with traditional loss functions, resulting in improved boundary detection

• Enhanced segmentation accuracy and generalization through post-processing and uncertainty-aware refinement techniques

Bias Detection in Political Discourse Using Deep NLP

• Designed and implemented a comparative framework for media bias classification using transformer models and a custom CNN-GRU architecture, analyzing both binary and multi-class political bias across prominent datasets

• Employed advanced text preprocessing, GloVe embeddings, and fine-tuning strategies to optimize semantic understanding while mitigating overfitting through regularization and early stopping techniques

• Surpassed prior benchmarks by achieving a 62.2% accuracy on the eleven-label classification task, demonstrating the model’s improved sensitivity to nuanced ideological bias across media sources

• Member of MSU Club Golf Team
• 5x Dean's List
• SpartaHack9 Winner

• Hindi: Native Proficiency
• English: Full Professional Proficiency
• Punjabi: Native Proficiency