Venkata Mahesh Kundurthi

Project: Real-Time AR-Enhanced Printer Troubleshooting Powered by Machine Learning

Key Achievements:

• End-to-End Model Development: Designed and deployed machine learning models to analyze real-world photo scenarios, achieving object detection accuracy above 85% in challenging conditions such as low-light, blurry, and high/low-resolution images.
• Data Augmentation: Enhanced model robustness through techniques like random cropping, rotation, and brightness adjustments, increasing training dataset size by 40%.
• Algorithm Optimization: Implemented the YOLO V6 algorithm for real-time object detection and annotation, achieving 6x faster processing speeds while maintaining high accuracy.
• Augmented Reality Integration: Enabled AR to identify and annotate printer and copier components accurately, reducing troubleshooting time by 50% and eliminating the need for on-site technicians.
• Collaboration and Quality Assurance: Collaborated with cross-functional teams, optimized algorithms for extensive datasets, and ensured compliance with strict data privacy standards.

Technologies Used: TensorFlow, PyTorch, YOLO V6, OpenCV, Python, ARKit/ARCore.

Project: Intelligent 5G Center Planning and Deployment Optimization

Key Achievements:

• Data Engineering: Developed scalable pipelines using Apache Kafka to process 15 TB+ data with 97.8% reliability; streamlined data retrieval with Google BigQuery, cutting latency by 30%.
• Predictive Modeling: Forecasted network demand with 91% accuracy, leveraging scikit-learn and reducing 5G deployment costs by 25% through optimized resource allocation.
• Anomaly Detection: Enhanced network reliability by identifying irregularities using Isolation Forest and Autoencoders, improving issue detection efficiency by 40%.
• Customer Insights: Boosted engagement by 15% through personalized marketing campaigns powered by advanced recommender systems.

Technologies Used: Apache Kafka, Hadoop HDFS, Amazon S3, Spark, scikit-learn, TensorFlow, Python.

Project: Email Spam Detection System

Key Achievements:

• Data Collection and Preprocessing: Gathered and cleaned 100k+ email datasets from various sources using Python and Pandas, ensuring 98% data quality by addressing missing values and balancing class distribution.
• Feature Engineering: Extracted key features such as word frequency, n-grams, and TF-IDF scores using scikit-learn, reducing irrelevant data by 30% and improving model interpretability.
• Model Development: Built and optimized machine learning models including Logistic Regression, Random Forest, and Naive Bayes, achieving a 90% spam detection accuracy.
• Bias Mitigation: Addressed data bias by applying SMOTE (Synthetic Minority Oversampling Technique) to balance classes, enhancing the model's notable recall improvement.
• Deployment: Deployed the system with Flask, enabling real-time email classification with a response time of Technologies Used: Python, Pandas, scikit-learn, Flask, SMOTE, TF-IDF.