Swapnil Bhattacharya

Data Scientist familiar with gathering, cleaning and organizing data for use by technical and non-technical personnel. Advanced understanding of statistical, algebraic and other analytical techniques. Highly organized, motivated and diligent with significant background in Machine Learning.

Generating Music Using LSTM Neural Networks. Link: https://shorturl.at/Wr340

• Developed and implemented a robust Long Short Term Memory (LSTM) neural network using Keras, achieving significant success in generating original music compositions from MDI files, demonstrating strong capabilities in sequence prediction and pattern recognition.
• Utilised the Music21 toolkit to efficiently parse and convert MDI files into musical notation, enabling precise handling of notes and chords, and facilitating seamless integration with the LSTM model for music generation.
• Conducted extensive training of the LSTM model with a dataset of piano music, implementing model checkpoints and dropout layers to prevent overfitting and enhance performance, resulting in the generation of coherent and aesthetically pleasing music pieces.

Lisp Interpreter. Link: https://shorturl.at/yHbXu

• Developed a comprehensive Lisp interpreter from scratch, employing a robust parser to handle Lisp’s unique syntax and leveraging evaluation and application function to process complex s-expression efficiently.
• Engineered the interpreter to manage built in functions and user defined constructs like lambda and if, ensuring accurate and context aware execution of Lisp code, thereby enhancing the interpreter’s versatility and reliability.
• Integrated efficient environment handling dynamically bind variables and evaluate function bodies, resulting in responsive and accurate interpretation of Lisp programs, while also facilitating seamless addition of new built-in-functions.

Content Based Recommendation Engine. Link: https://shorturl.at/S0oGS

• Developed a robust recommendation engine using Python’s Scikit Learn library, implementing TF-IDF and cosine similarity

     to analyse project description and generate highly relevant recommendations, enhancing user experience on eCommerce

     platforms.

• Utilized Redis for efficient storage and retrieval of similarity scores, enabling rapid real-time recommendations by precomputing and caching the top 100 most similar items for each product, significantly improving system performance.
• Successfully created and deployed a content-based recommendation engine on Heroku, leveraging a Flask framework to serve recommendations via a REST API, ensuring easy integration and scalability for real-world eCommerce applications.