Experience

• Leading a 3-member research team, including 2 undergraduate researchers, to design and validate NeuroSwin, a hybrid Swin Transformer-GRU model for Parkinson's Disease classification and brain-region localization from EEG signals.

• Co-developed TransfusionNet, a hybrid architecture with a specialized preprocessing pipeline and early-layer feature fusion for cervical cancer detection. Published in the Q1 journal Results in Engineering (Elsevier, Impact Factor: 9.4).
• Created a benchmark to evaluate the performance of large language models (LLMs) in identifying disease classes, comparing results against our proposed architecture MobileCoAtNet for stomach infection classification.
• Investigated the adaptability of foundational AI models for multi-cancer classification, implementing Model-Agnostic Meta-Learning (MAML) for domain adaptation to enhance generalizability.
• Actively researching the feasibility of large language models (LLMs) for disease diagnosis, reasoning, and treatment follow-up in medical applications.
• Designed SALT, a spiking neural decoder that uses a state-space model to adaptively set firing thresholds, establishing a causal accuracy-energy Pareto benchmark across six architecture families for closed-loop brain-computer interfaces; under review in Neural Networks.
• Developed CT-SALT, a cross-timescale extension of SALT in which a slow population branch gates a fast spiking branch's firing threshold, improving decoding accuracy and training stability at no added energy cost.

• Independently conceived, implemented, and authored a solo study on gallbladder cancer classification using ensemble learning, published at ICEEICT 2024.
• Designed a hybrid model for gallbladder cancer classification using parallel transfer learning, feature fusion, and LSTM as the undergraduate thesis, of which the ICEEICT 2024 paper formed a portion; the full study is in preparation for Expert Systems with Applications.
• Conducted the literature review supporting the feature-fusion approach, drafted portions of the thesis manuscript, handled data preprocessing, and presented the completed work at the final thesis defense.

Khulna, Bangladesh

Presented Paper: Exploring Common Molecular Interactions across Multiple Cancers to Identify Potential Therapeutic Targets and Drug Candidates

Type: Oral Presentation

Role: Presenter

Department of Biomedical Engineering, Jashore University of Science & Technology, Jashore, Bangladesh

Presented Poster: Exploring Gene Network Interactions and Pathways in Three Cancers: Identifying Potential Therapeutic Targets and Drug Candidates

Type: Poster Presentation

Role: Presenter

Dhaka, Bangladesh

Presented Paper: Classification of Gallbladder Cancer Using Average Ensemble Learning

Type: Oral Presentation

Role: Presenter