Mr. Md Zafar Imam

Assistant Professor

Mr. Md Zafar Imam joined the Department of Computer Science and Engineering at Dayananda Sagar University, Bengaluru, as Assistant Professor. His work centers on applying deep learning to two problem spaces: computer-vision-based diagnostic systems and privacy-preserving natural language processing for sensitive, real-world data.

One strand of his work addresses plant disease detection in agriculture. He built a deep learning model using ResNet50 and EfficientNet (transfer learning) on the PlantVillage dataset to classify 38+ crop diseases across 14 plant species from leaf images. Applying image augmentation such as rotation, flip, zoom, and shear alongside dropout regularization, the model reached over 96% validation accuracy. He then deployed the system end-to-end as a Streamlit web application, allowing users to upload a leaf image and receive a disease diagnosis along with suggested remedies in real time.

On the medical imaging side, he developed a hybrid deep learning framework, Hybrid-MSGFNet, for 4-class brain tumor classification from MRI scans. The model combines DenseNet-121, CBAM, a Transformer-based global context encoder, cross-attention, multi-scale refinement, and gated residual fusion. To prevent data leakage across 7,023 MRI images, he implemented MD5 and perceptual hashing with group-wise splitting, and the framework achieved 98.82% test accuracy, a 0.99 macro-F1 score, and a 0.99 AUC using test-time augmentation.

His current research is directed toward privacy-preserving natural language processing, aiming to develop secure and trustworthy Large Language Models (LLMs) using Federated Learning, Differential Privacy, and privacy-aware fine-tuning techniques. His ongoing project, Federated Fine-Tuning of Large Language Models for Privacy-Preserving Natural Language Processing, investigates how LLMs can be fine-tuned across multiple decentralized data sources using Federated Learning. By integrating Differential Privacy with Parameter-Efficient Fine-Tuning (PEFT) methods such as LoRA, the framework enables collaborative model training while ensuring that sensitive textual data remains on local devices or organizational servers, without requiring centralized data collection. The goal is to build secure, scalable NLP systems suitable for healthcare, finance, legal, and enterprise applications.

He holds an M.Tech in Computer Science and Engineering from NIT Silchar and a B.Tech in Computer Science and Engineering from Nalanda College of Engineering, Chandi, and is comfortable across the ML/DL stack: Python, C++, C, and SQL, with hands-on experience in deep learning, NLP, and object-oriented programming.

Before moving into academia, he built a MERN-based Academic Record Management System with PostgreSQL, featuring secure REST APIs, JWT authentication, and a responsive React dashboard for managing student records, attendance, and performance. He is now looking to build out both of his ongoing research threads, privacy-preserving NLP and applied computer vision, into a sustained research programme at DSU.

Email: This email address is being protected from spambots. You need JavaScript enabled to view it.

LinkedIn: https://www.linkedin.com/in/md-zafar-imam-8940901b2/

GoogleScholar: https://scholar.google.com/citations?hl=en&user=P1mD3nYAAAAJ