Ananya Mantravadi

Graded assignments, evaluated projects, mentored, and conducted tutorial sessions for 7 graduate and undergraduate courses like Object-Oriented Design and Development, Operating Systems, and Data Structures. Guided over 400 students across 3 institutions.

Developed and deployed an MLOps pipeline for revenue forecasting using time-series modeling with Python, Pandas, and DataRobot, integrating product capacity and usage data, CI/CD workflows, and quarterly backtesting (90/60/30-day rolling updates), reducing forecasting error to 1% (MAPE) for a $30M business unit; presented insights and visualizations (Tableau) to business stakeholders.

Built and evaluated customer churn prediction models for a SaaS product using XGBoost, Dynamic Time Warping, Tableau, and survival analysis; estimated potential loss and retention ROI, uncovered data limitations that informed strategic pivots.

Implemented and enhanced image dehazing, PCB components inspection, dental X-ray segmentation, and arrhythmia classification using novel neural networks and deep learning techniques, leading to scholarly publications.

Designed, built, and deployed a voting ensemble classification model using Python, SQL, and Azure ML for predicting cases likely to default while extending loan offers for Scorpio Z101, a newly launched SUV improving recall by 30% to achieve 72%. This model was developed using features extracted from large-scale credit bureau data and the internal bank database.

Developed and executed ETL pipelines, utilizing Python and SQL, to process retro scrub data from diverse credit bureau sources such as CRIF, CIBIL, and Experian, reducing data processing time by 20 hours.

Developed an interactive stock analysis dashboard using Python, Dash/Plotly, CSS, & Yahoo Finance API, integrating real-time data, user interaction logic, and dynamic Ul elements for technical indicators, risk modeling, and future projections (GBM, GARCH). (Link)