- Resumes: https://github.com/i-majumder/Resume-tamal
- LinkedIn Profile: https://www.linkedin.com/in/tamal-majumder/
- Master of Science (MSc) in Physics, Indian Institute of Technology (IIT Delhi) 🏛️, 2023-2025
- Bachelor of Science (BSc) in Physics Honours, Asutosh College, Calcutta University 🏫, 2020-2023
- Machine Learning 📚: Supervised Learning • Unsupervised Learning • Model Deployment 🚀 • Model Optimization ⚙️ • Hyperparameter Tuning • Model Interpretability (SHAP) 🧐 • AutoML (H2O) 🤖 • Scikit-Learn • Classical Algorithms • Tree-Based Algorithms 🌲 • Ensemble-based Algorithms.
- Natural Language Processing (NLP) 📝: Recurrent Neural Network (RNN) • Hugging Face Transformers 🤗 • Text Classification • Text Generation • Text Summarization 📑 • Topic Modeling • Machine Translation 🌍 • Question Answering 💬 • NLP Data Preprocessing 🧹 • Generative AI with LLMs • Image-Text Multi-models 🖼️📝 • Semantic Search 🔍 • Vector Database • Named Entity Extraction.
- Computer Vision 🖼️: Image Processing • Convolutional Neural Networks (CNN) 🧠📷 • Image Segmentation • Object Detection 🕵️ • Image Classification • Transfer Learning 🔄 • Feature Extraction • Image Semantic Search 🔍🖼️.
- Model Deployment Tools 🌐: Streamlit • Flask • FastAPI 🚀 • Gradio 📡 • Heroku • Docker Containerization 🐳 • HTML and CSS 🎨.
- Additional Skills : Time Series Forecasting ⏳📈• Web Scraping (BeautifulSoup, Selenium, Requests) 🕸️ • Linux.
- Deep Learning Tools/Framework : NumPy • Pandas 🐼 • Scikit-Learn • TensorFlow • PyTorch • Keras • Hugging Face Transformers 🤗 • NLTK • SpaCy • Gensim • Word2Vec • GloVe
- Data Analysis and Visualization 📊: Data Wrangling 🧹 • Tableau • Plotly • Seaborn • Matplotlib .
- MSc Thesis - Deep Learning for DHM Phase image Reconstruction , Segmentation and Classifcation under (Prof. Kedar Khare)
Objective: Employ data analysis, simulation, and machine learning to Reconstruct DHM Phase Images and contribute to advancements in Biomedical Image Processing.
- Analyzed 1000 (512 x 512 px) high resolution phase images of Blood sample from 1000 individuals.
- Engineered a U-Net model for reconstruction and segmentation of blood sample images, after Data Augmentation.
- Trained several DNN models and choose ResNet-50 to classify healthy cells vs cells from a smokers blood sample.
- Achieved MSE of 0.0085 for reconstruction and for classifcation achieved accuracy of 91.2% and F1 score of 0.932
