Aasim Pasha

• Led a 12-member cohort through the full software development lifecycle - requirements analysis, technical design,
coding, and testing - writing clean, well-documented Python for an end-to-end ML pipeline, including an
Optuna-tuned XGBoost model (R2 > 0.92, MAE < 22 units) that outperformed physics-based estimation by 63%.
• Designed a modular, object-oriented billing engine covering all three Gruha Jyoti billing scenarios, applying secure
coding and software design best practices, conducted peer code reviews and collaborated cross-functionally on
technical design decisions.
• Debugged and optimized the pipeline for accuracy and performance, adding SHAP-based explainability and a
personalized appliance-level savings recommendation engine (units and INR).