ALL WORK APPLIED MACHINE LEARNING & DATA SCIENCE
SELECTED WORKS

FEATURED PROJECTS.

Curated portfolio of comparative forecasting models, natural language processing pipelines, and transformer recommendation architectures.
TIME-SERIES FORECASTING JULY 2025
01
ECONOMETRIC VS NEURAL MODELING

ARIMA vs LSTM — Stock Price Prediction

Built and evaluated statistical ARIMA against sequential LSTM networks for stock price movement and volatility forecasting on Tesla (TSLA) historical data via Yahoo Finance API.

PYTHON PYTORCH STATSMODELS STREAMLIT YAHOO FINANCE API
VIEW CASE STUDY → GITHUB ↗
Model Performance LSTM ACHIEVED LOWER RMSE
Evaluation Metrics

Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE).

NLP & CLASSIFICATION JUNE 2025
02
SENTIMENT PIPELINE

Understanding the Voice of Twitter

Complete NLP classification pipeline on Kaggle Twitter data using regex text normalization, WordNet lemmatization, and TF-IDF (unigrams + bigrams across 5,000 features) feeding regularized Logistic Regression.

PYTHON SCIKIT-LEARN NLTK TF-IDF LOGISTIC REGRESSION
VIEW CASE STUDY → GITHUB ↗
Accuracy Metric ~70% TEST ACCURACY
Dataset Specs

Multi-class Twitter sentiment corpus with positive, neutral, and negative classification targets.

NEURAL RECOMMENDERS MAR–MAY 2025
03
TRANSFORMER RECOMMENDATIONS

Game Recommendation Engine

Content-based recommender trained on multifaceted video game metadata. Uses BERT representations to capture deep semantic affinity and vector cosine similarity for ranking recommendation candidates.

PYTHON HUGGING FACE BERT COSINE SIMILARITY STREAMLIT
VIEW CASE STUDY → GITHUB ↗
User Engagement ~60% HIT RATIO (TOP-K)
Catalog Intelligence

Metadata inputs include user reviews, tags, plot summaries, genres, and developer studios.