Vasan Janakiraman Ramesh

Data Science graduate student with expertise in Statistical modeling and time series analysis with experience in machine learning, deep learning, and database management. Proficient in Python, R, SQL, Snowflake, and Tableau with many projects using LSTM networks, NLP systems and other regression models with real-world datasets including Lahman Baseball Database for sports analytics and Twitter text data for text analysis. Seeking data science internship to apply my skills to real-world challenges.

StockVision:Predicting Market Trends with LSTM and Moving Averages, Developed a predictive system utilising an Ensemble Machine Learning model based on historical data., Integrated Long Short-Term Memory (LSTM) networks and Moving Average techniques to enhance forecasting capabilities., Achieved improved accuracy and reliability through insights gained from domain experts., Implemented real-time data ingestion to keep the model updated with the latest historical data. 

HateMonitor:Identifying Offensive Language, Engineered a hate speech detection system leveraging Natural Language Processing (NLP) and machine learning techniques specifically designed for monitoring and analyzing discussions in college forums, promoting a safer online environment., Used Sentiment Analysis to accurately identify harmful and offensive language in text data., Worked with real-world datasets to enhance the model's effectiveness and reliability.

Pharmaceutical Cross-Series Demand Forecasting Project:Engineered a robust forecasting framework for eight pharmaceutical product categories using advanced statistical models (ARIMA, SARIMA, GARCH) and multivariate time series analysis, Designed and validated a cross-series approach that identified N02BE (OTC analgesics) as a leading indicator, improving forecast accuracy for related categories by up to 26.6%,Employed Vector Autoregression (VAR), Granger causality testing, and Impulse Response Function (IRF) analysis to rigorously uncover dynamic dependencies between series and quantify predictive relationships, Developed and tuned predictive models that delivered industry-relevant metrics (e.g., MAPE reduced from 56.8% to 41.7%), enabling early-warning inventory strategies and reduction of stockouts, Collaborated on full pipeline: data wrangling, exploratory analysis, statistical testing, model deployment, and executive-level reporting; ensured findings were actionable for supply chain and inventory management.