Janhavi Doshi

Projects

Real-Time Financial News-Driven Stock Prediction Web Application

  • Led a multidisciplinary team of 5 in SQL database design and integration of BERT large language model for sentiment analysis.
  • Developed a machine learning model to predict stock price direction by real-time financial news sentiment analysis.
  • Built a user-friendly web application using Flask that suggests the users whether to buy/ sell the selected stock.

Custom Database Enhancements - H2 Database

  • Developed a custom aggregate function to identify the nth maximum value, enhancing data retrieval efficiency.
  • Designed a specialized 'Contact' datatype for H2 database, ensuring validation of contact details like phone numbers and emails.
  • Updated Java H2 in-memory database files, adding native support for the custom function and datatype to enhance performance.

Language Classifier AI

  • Developed a Python-based machine learning model proficient in differentiating between English and Dutch languages.
  • Employed a combination of decision tree and random forest algorithms for accurate test classification.
  • Crafted comprehensive unit tests and debugging protocols, achieving a 95% accuracy in model performance.