Prakhar Mishra

• Designed and implemented a Natural Language Generation (NLG) model to automatically generate concise and coherent summaries from any given topic.

• Applied text preprocessing, NLP techniques, and sequence-to-sequence architectures to improve output accuracy and readability.

• Conducted experiments to optimize model performance, reducing redundancy and enhancing summary quality.

• Collaborated with a team to integrate the model into a real-world application, gaining hands-on experience in AI/ML deployment.

Pathfinder for UPES (Minor Project)

• Developed a campus navigation tool to assist students and visitors in locating classrooms, labs, and facilities.

• Implemented using C++, Node.js, and HTML/CSS for interactive and user-friendly design.

• Reduced navigation time by 30% through optimized pathfinding algorithms.

• Integrated real-time updates for events and construction zones to improve usability.

Movie Recommender System

• Built a content-based movie recommendation engine to suggest personalized movies to users.

• Utilized Python, Scikit-learn, Pandas, and NumPy for data preprocessing and model building.

• Implemented TF-IDF vectorization to analyze and compare movie metadata.

• Deployed locally via Flask, achieving 85% accuracy in recommendations.

Vision - OCR Powered | PyTorch, Kaggle, Jupyter Notebook

• Developed an OCR system to extract and recognize text from scanned images and documents.

Converted images/PDFs to machine-readable text using EasyOCR.
• Used BART-large MNLI model to determine whether text is AI- or human-generated.

• Implemented using Python, OpenCV, and Tesseract OCR for accurate text detection.

• Enabled conversion of unstructured image data into editable and searchable text for practical use cases.