PyTorch & Deep Learning | Python Developer
Engineered a custom convolutional neural network with PyTorch to accurately classify rice varieties. Designed efficient data preprocessing and augmentation pipelines, and implemented GPU-accelerated training for rapid experimentation and model optimization.
Designed and developed a smart chatbot in Python by integrating Google's Gemini AI model through API calls. Focused on user experience with a clean interface and demonstrated mastery of external API integration and custom chatbot logic.
Built an end-to-end object detection pipeline to locate vehicles in images. Performed dataset annotation, model selection, and training, with a strong focus on accurate localization through bounding box visualization and post-processing.
Developed a predictive model to estimate Titanic passenger survival probabilities using feature engineering and statistical modeling. Applied domain expertise to preprocess data and selected optimal classifiers for robust performance.
Built a deep CNN in PyTorch for multi-class animal face classification using the AFHQ dataset. Leveraged data augmentation, advanced preprocessing, and model regularization to achieve high accuracy and strong generalization.
Created PersonalityPulse, a personality prediction web app backed by a PyTorch classification model. Achieved ~99% accuracy on Kaggle data, and enabled live predictions using a fast, user-friendly Streamlit interface.
Developed an end-to-end brain tumor detection pipeline using PyTorch, leveraging ResNet18 with U-Net-inspired segmentation and Grad-CAM visualizations for model explainability. Achieved strong accuracy on public MRI datasets, focusing on practical medical AI applications.
Developed a rule-based NLP chatbot in Python, using TextBlob for sentiment analysis and dynamic response tailoring. Demonstrated understanding of conversational AI and natural language processing workflows.
Sep 2023 - Present
May 2023 - Sep 2023
Scalable image detection in the cloud—PyTorch + AWS Lambda. In Progress
CNN-based disease diagnosis and severity scoring for agriculture efficiency.
Developing an LSTM-based model for real-world sentiment classification tasks.
Using regression models to forecast bike rentals based on temporal and weather data.
Harnessing RNNs to generate creative television scripts from sample dialog data.
Training GANs (Generative Adversarial Networks) to synthesize realistic human faces.
End-to-end deployment of a sentiment analysis model as a web API with live demo.