Text Sentiment & Stress Detection
Fine-tuned BERT on Reddit/Twitter data for stress classification. F1 = 0.88, ROC-AUC = 0.94 — outperformed RNN and Transformer baselines.
view on github →Fine-tuned BERT on Reddit/Twitter data for stress classification. F1 = 0.88, ROC-AUC = 0.94 — outperformed RNN and Transformer baselines.
view on github →1,982-image custom annotated dataset. Two-stage YOLOv11 + YOLOv8-cls pipeline. 0.941 precision in binary detection. Synthetic augmentation via Blender.
view on github →GAN-generated synthetic defect images integrated into CNN training. Reduced false negatives and improved defect class recall by 22%.
view on github →Star-schema dimensional model with Apache Hop ETL into Oracle Autonomous DB. Power BI dashboards for operational KPI tracking.
view on github →Interactive Tableau dashboard for customer behavior analysis — seasonality, segmentation, and demand forecasting for non-technical stakeholders.
view on tableau →Logistic Regression, Random Forest, XGBoost with PCA + cross-validation. Deployed via FastAPI + Docker for real-time prediction endpoints.
view on github →Google Gemini NLP parsing + Google Sheets pipeline. Reduced manual data entry by 96% — from 5 minutes to 10 seconds per record.
view on github →3NF relational schema + Python ETL across 10K+ records. Indexing improved query performance by 65%. Streamlit dashboards for decision support.
view on github →