Ananya
Barthakur

A React web application that combines computer vision and NLP to help users identify plants and receive personalized care guidance. The app features dual AI integration with offline fallback capabilities.

Key Features

• ▸Smart Plant Identification: Integrated OpenAI Vision API and PlantNet API with fallback logic, achieving 90%+ accuracy in plant identification
• ▸AI-Powered Chat Assistant: Implemented conversational AI using OpenAI's GPT-4 for personalized plant care advice and problem diagnosis
• ▸Robust Architecture: Built comprehensive offline mode with local plant database, ensuring 100% functionality even when AI services are unavailable
• ▸User Experience: Designed intuitive interface with real-time image analysis, confidence scoring, and detailed care instructions
• ▸Performance Optimization: Implemented retry logic, error handling, and graceful degradation for enterprise-level reliability

Python implementation for electronic energy approximation and molecular orbital visualization in quantum chemistry applications.

Key Features

• ▸Implemented Self-Consistent Field (SCF) algorithm for solving the many-electron Schrödinger equation
• ▸Built efficient integral evaluation and matrix manipulation routines for molecular systems
• ▸Achieved numerical stability through symmetric orthogonalization and convergence acceleration
• ▸Designed modular architecture for different basis sets and molecular configurations

Developed CNN and RNN models for breast cancer tissue analysis, enabling automated identification of tumor characteristics in TNBC research.

Key Features

• ▸Implemented U-Net and FCN architectures for nuclei segmentation in histopathological images
• ▸Achieved greater than 90% accuracy in identifying cancer nuclei morphology and characteristics
• ▸Built comprehensive data augmentation pipeline for medical imaging robustness
• ▸Designed custom loss functions combining Dice loss and boundary loss for precise segmentation

Built Kernel Ridge Regression model with Gaussian Processes to predict molecular potential energy surfaces, using PyTorch for automatic differentiation.

Key Features

• ▸Implemented Kernel Ridge Regression with custom kernel functions for molecular property prediction
• ▸Utilized Gaussian Processes for uncertainty quantification in energy surface mapping
• ▸Leveraged PyTorch's automatic differentiation for gradient-based optimization
• ▸Achieved high accuracy in predicting molecular configurations and energy landscapes

Django web application for Boston College streamlining Teaching Assistant applications with secure authentication and administrative workflows.

Key Features

• ▸Built full-stack web application using Django framework with PostgreSQL database
• ▸Implemented secure user authentication and role-based access control
• ▸Designed administrative dashboard for faculty to review and manage TA applications
• ▸Created responsive frontend with dynamic forms and real-time application status updates