Othoniel Joseph
Skills
PROGRAMMING LANUAGES: PYTHON, C++, JAVA
DATA SCIENCE: MACHINE LEARNING, PREDICTIVE ANALYTICS, GENERATIVE AI (RAG), DATA MODELING
VISUALIZATION: MATPLOTLIB, SEABORN, PLOTLY, RSTUDIO, TABLEAU
TOOLS/FRAMEWORKS: KERAS, TENSORFLOW, PYTORCH, CURSORAI, VSCODE, VISUAL STUDIO
DATABASES: SQL, POSTGRESQL
SOFT SKILLS: LEADERSHIP, TEAM COLLABORATION, STAKEHOLDER COMMUNICATION
PROJECTS
INDEPENDENT STUDY ~ DEEP LEARNING AND BUSINESS ANALYTICS
Developed a Spatio-Temporal Graph Neural Network (ST-GNN) to forecast semiconductor stock prices using financial indicators, technical metrics, and FinBERT-based sentiment from over 370 days of financial news. Achieved a MAPE of 3.2% for large-cap stocks. Engineered dynamic graph structures with multi-edge types (price, RSI, MACD, sentiment) to capture evolving financial relationships. Reduced validation loss by 94%, and visualized model performance using RMSE, R², and feature importance rankings. Built a full ML pipeline with custom Python scripts for data collection, preprocessing, graph construction, training, and evaluation.
AI & BUSINESS SOLUTIONS ~ FULL STACK RAG SYSTEM
Led a team of 4 to build a full-stack AI-powered application using Python, Supabase, OpenAI API, FastAPI, Next.js, and Tailwind CSS. Conducted stakeholder interviews to align technical solutions with business goals. Designed and implemented a Retrieval-Augmented Generation (RAG) system tailored to real-world client use cases. Delivered weekly progress reports and a flawless live demo. Accelerated development cycles by 30% using Cursor AI, GitHub Copilot, and Replit.
CAPSTONE PROJECT ~ SONG POPULARITY CLASSIFIER
Solved a classification problem to predict a song's popularity using a dataset of 600 records and 21 aggregated features. Performed data preprocessing, visualization, feature selection, and logistic regression modeling. Achieved 85% accuracy. Built a full data science pipeline from feature engineering through model evaluation to deliver actionable insights.
MACHINE LEARNING ~ STOCK FORECASTING WITH LSTM
Led a group project analyzing 100,000+ airline delay records to evaluate their impact on stock prices. Conducted EDA and correlation analysis in RStudio, developed a linear regression model, and achieved an R² of 0.85. Delivered insights into delay-stock price relationships through visualizations.
DATA MINING ~ MODEL COMPARISON AND OPTIMIZATION
Compared classification algorithms including KNN, SVM, Random Forest, Decision Tree, and XGBoost on benchmark datasets. Applied feature extraction and engineering techniques, improving model performance by 15%.
DATA SCIENCE AND BUSINESS ANALYTICS
Developed a data science pipeline for a real-time business application using Python, SQL, and Power BI. Achieved a 20% reduction in data processing time and a 15% improvement in model accuracy. Demonstrated ability to translate business requirements into data-driven solutions.