+ Conducted Lectures: Independently designed and lead engaging STAT101A classes that complement the foundational concepts of Statistics, resulting in a 89% pass rate among students+ Interactive and Practical Learning: Cultivate an interactive learning environment by providing hands-on learning experiences and assignments that reinforce theoretical knowledge and enhance practical skills, encouraging active participation and fostering a deeper understanding of statistical principles.+ Performance Analysis: Analyzed class performance data to identify trends and improve teaching strategies

+ Course Audit and Evaluation: Monitored student engagement, designed/published surveys for PHIT 306, analyzed feedback to generate comprehensive reports and recommendations for the university, aiding in informed decision-making regarding the course’s continuation and improvements+ Data Pipeline Creation: Engineered and implemented a comprehensive data pipeline for university food pantry supplies, optimizing data collection, warehousing, EDA, predictive analytics, and visualization to improve supply management.+ Achievement Disparity Assessment: Assessed achievement disparities among high school students across various factors and formulated actionable strategies for optimizing year-on-year GPA and enrollment+ Enhancing Education Quality: Contributed to the university’s mission of delivering high-quality education by providing valuable insights and recommendations for curriculum refinement

+ Engineered a robust Faster R-CNN model for clothing object detection, achieving a weighted average F1 score of 0.700.+ Constructed a multitask Siamese ResNet50 to calculate embeddings for image similarity, yielding an average cosine similarity of 0.901 for related products, thereby effectively driving recommendations for analogous clothing selections.+ Apply K-means and hierarchical clustering on multi-dimensional customer purchase data for audience segmentation.+ Transitioned research endeavors above into production-ready codebases, ensuring systematic development processes.

+ Payment stoppage detection algorithm (anomaly detection) : Created a process to filter out customers who stopped paying to the client using past due amount buildup and z-score. Improved model accuracy by removing the outlier customers+ Runtime and Memory Optimization: Optimized runtime and CPU utilization of ERP Data based Cash forecasting framework by 15% and 40% respectively, effectively improving implementation efficiency and reducing operation costs on 10%, on datasets with ~20M records by leveraging multiprocessing and data aggregation.+ Cash Forecast and Variance Explainability Analytics: Conceptualized and productized analytical reports for clients to use to understand reasons for variances in the cash forecast at different component and granularity levels+ Unsupervised Clustering : Executed POC for unsupervised bank transaction classification into cash flow category using tokenized vectors, clustering algorithms - K-means, DBSCAN, Hierarchical and employing approximate subset sum using Dynamic Programming (DP) with solution backtracking, to bring about 85% amount match against aggregate GL data.

+ Model Migration: Spearheaded large-scale model migration effort, collaborating with 50+ trainees to transfer, test, and deploy legacy predictive models to a newer architecture, achieving a 99% success rate and reducing deployment time by 30%.+ Model Benchmarking: Conducted comprehensive analysis of various neural networks (Vertex AI, Abacus AI, ARIMA+, Bi-LSTM, FB-Prophet) to identify top-performing models for cash forecasting, resulting in a 15% accuracy improvement over existing regression models and informing strategic upgrade decisions.+ Document Classification Expertise: Designed and developed a CNN-based model for scanned financial document classification, streamlining remittance processing with 92% accuracy and reducing manual processing time by 40%.+ Forecasting Framework Development: Created an innovative Invoice Payment Delay Prediction framework, leveraging Regression models (Linear, Bagging, Boosting) for accurate forecasting (~100 weekly) with a 12% increase in forecast precision, enabling proactive Collections Analyst optimization of Days Sales Outstanding and reducing outstanding invoices by 8%.+ Quality Assurance : Ensured robust code quality by building a comprehensive Pytest testing suite, achieving 95% code coverage across forecasting modules and utilities, and reducing bugs by 20%.

+ Medical Image Analysis : Implemented deep learning models for analyzing medical images (X-rays) to assist in diagnosing breast cancer.+ Documentation : Made a detailed report covering the Theory, Methodology and Implementation of the project and visualized the problems+ Healthcare Performance Analytics : Built a dashboard that tracks and analyzes their performance metrics, patient outcomes, and operational efficiency.+ Health Risk Assessment Tool : Created an application that assesses an individual's health risk based on their lifestyle, medical history, and other personal data.