Venkata Mahesh Mopidevi

AI Software Engineer with an MSc in Data Science and hands-on experience designing production-oriented machine learning and computer vision systems. Experienced in developing end-to-end AI pipelines using Python, OpenCV, MediaPipe, YOLO, PyTorch, FastAPI, Docker, MLflow and GitHub Actions. Built AI solutions ranging from real-time biomechanical analysis using computer vision to production-ready ML platforms with automated deployment workflows. Passionate about applying AI research to solve real-world engineering challenges through clean, scalable software.

AI Cricket Coach — Computer Vision Batting Analysis Platform, Python

OpenCV, MediaPipe, YOLO, PyTorch, Computer Vision 

A computer vision platform that transforms smartphone cricket videos into automated biomechanical analysis and coaching feedback using pose estimation and object detection.

• Built an end-to-end AI pipeline that converts batting videos into pose analysis, impact detection, biomechanical metrics and automated coaching reports.
• Integrated MediaPipe Pose (33 landmarks) with custom YOLO object detection models trained on 2,700+ labelled bat, ball and batter images.
• Designed a hybrid dual-YOLO detection pipeline achieving 93–100% batter detection and up to 87% accepted bat detection on real batting footage.
• Developed six biomechanical batting metrics, automated shot-phase detection and evidence-based coaching reports with prioritised improvement recommendations.
• Engineered a modular Python architecture comprising 58 source modules, 14 command-line analysis tools, and 76 automated unit tests.
• Implemented multi-stage ball tracking and temporal validation to reduce false positives and improve bat–ball impact detection reliability.
• Built a human-in-the-loop validation framework supporting coach verification and future model calibration.

Technologies: Python, Ultralytics YOLO, Pandas, NumPy, Matplotlib, Git, Pytest.

AI Banking Risk Intelligence Platform, Python

FastAPI, MLflow, Docker, GitHub Actions.

Designed and developed a production-oriented machine learning platform for financial risk prediction using modern MLOps practices.

• Built an end-to-end machine learning pipeline covering preprocessing, feature engineering, training, evaluation and deployment.
• Achieved approximately 83% classification accuracy with 0.93 ROC-AUC using ensemble learning models.
• Developed FastAPI inference services exposing production-ready prediction endpoints.
• Integrated MLflow Model Registry for experiment tracking, model versioning and deployment management.
• Containerised the platform using Docker and Docker Compose.
• Automated testing and deployment using GitHub Actions CI/CD workflows.
• Designed modular software architecture enabling future retraining and scalable deployment.

Technologies: Python, Scikit-learn, GitHub Actions.

ShiftGoal (iOS Application)

Designed and developed an iOS productivity application for shift workers to track working hours, earnings and personal goals using SwiftUI. Published for the Apple ecosystem with modern mobile development practices.

• MSc Data Science
• Level 2 Food Safety & Hygiene