R Srinivas Prabhu

Offering solid foundation in data analysis and machine learning with keen desire to learn and grow in field. Delivers foundational understanding of Python, aiming to build upon this with advanced techniques and tools. Ready to use and develop data analysis and machine learning skills in Data Scientist role.

Hit-and-Run Prediction (Research Paper | Ensemble Models | SMOTE)

• Compare Model accuracy for imbalanced datasets with / without implementation of SMOTE.
• Conducted a comparative analysis of 2 ensemble models (Random Forest,XGBoost) on 25,000+ real-time traffic crash records from the Chicago Data Portal.
• Applied SMOTE, improving recall and overall accuracy by 7–10%, and reducing false negatives in hit-and-run classification.
• Authored and presented research findings at the Microsoft CMT Conference, Egypt, showcasing the impact of SMOTE on model performance.

Demand Forecasting Dashboard (Time series | Demand forecasting | Timesfm )

• Built a cloud-native forecasting system with FastAPI (REST API) and Streamlit (dashboard), enabling real-time demand predictions with .
• Trained and deployed XGBoost and TimesFM (LLM) models, improving forecast accuracy by ~18% MAPE reduction vs baseline.
• Automated end-to-end ML pipeline (data preprocessing, training, deployment) via GitHub Actions + Docker + GKE, cutting deployment time from hours to .
• Scaled services on Kubernetes with health checks and Ingress, supporting 1k+ concurrent requests without degradation.

DeepGLO – Graph-based Time Series Forecasting Library

• Developed a Python library for deep graph learning + time series forecasting, integrating matrix factorization for global patterns with Temporal Convolutional Networks (TCN) for local temporal modeling.
• Delivered a scikit-learn–compatible API with modular, extensible design, enabling seamless adoption in ML pipelines.
• Achieved state-of-the-art forecasting accuracy on multivariate time series benchmarks by jointly modeling inter-series correlations and temporal dependencies.
• Optimized for research & production use, supporting scalable training and flexible experimentation across large datasets.

IBM Data Analyst Certificate

• Mastered data cleaning, visualization, SQL, Excel, Python, and dashboard creation. Applied analytical skills to deliver actionable insights.