Shreyansh Chopra

<p>Developed a machine learning pipeline in Python, employing boosting techniques to enhance model accuracy.<br>
Addressed class imbalance using advanced sampling techniques, including SMOTE and Tomek Links, increasing model precision from 76% to 85% and reducing engine test failures by 15%.<br>
Containerized the pipeline using Docker and leveraged AWS Lambda to process real-time data from PostgreSQL triggers, enabling consistent, scalable deployments.<br>
Designed and deployed an AI-powered multilingual sales enablement platform using AutoGen agent and Azure AI, integrating NLP to automate personalized pitch generation tailored to regional languages.<br>
Utilized a Neo4j knowledge graph database to structure and retrieve interconnected data on regions, retailers, incentive schemes, and historical performance for intelligent pitch customization.<br>
Achieved a 40% reduction in pitch preparation time, accelerating sales team responsiveness.<br>
Engineered a real-time defect detection system for mechanical dies using YOLOv8, a cutting-edge object detection model in computer vision, achieving 90% accuracy for precise defect localization in manufacturing.<br>
Built and deployed an end-to-end data pipeline by annotating 5,000 die images using LabelImg, applying OpenCV for image preprocessing, and storing data in a structured format for efficient model fine-tuning.<br>
Reduced sheet metal part rejections by 20–30%, minimizing material wastage and improving efficiency.</p>

Engine Testing Classification
• Developed a machine learning pipeline in Python, employing boosting techniques to enhance model accuracy.
• Addressed class imbalance using advanced sampling techniques, including SMOTE and Tomek Links,
increasing model precision from 76% to 85% and reducing engine test failures by 15%
• Containerized the pipeline using Docker and leveraged AWS Lambda to process real-time data from
PostgreSQL triggers, enabling consistent, scalable deployments.
Bazaar Sales Optimization
• Designed and deployed an AI-powered multilingual sales enablement platform using AutoGen agent and
Azure AI, integrating NLP to automate personalized pitch generation tailored to regional languages.
• Utilized a Neo4j knowledge graph database to structure and retrieve interconnected data on regions, retailers,
incentive schemes, and historical performance for intelligent pitch customization.
• Achieved a 40% reduction in pitch preparation time, accelerating sales team responsiveness.
Defect Detection in Automotive Manufacturing
• Engineered a real-time defect detection system for mechanical dies using YOLOv8, a cutting-edge object
detection model in computer vision, achieving 90% accuracy for precise defect localization in manufacturing.
• Built and deployed an end-to-end data pipeline by annotating 5,000 die images using LabelImg, applying
OpenCV for image preprocessing, and storing data in a structured format for efficient model fine-tuning.
• Reduced sheet metal part rejections by 20–30%, minimizing material wastage and improving efficiency

<p>Developed a robust Speech-to-Text (STT) conversion module using Whisper AI for high-accuracy transcription, integrating advanced Natural Language Processing (NLP) techniques to uncover customer interaction insights.<br>
Applied sentiment analysis and emotion detection algorithms alongside frequency analysis to classify customer sentiments and key customer grievances and provided strategic recommendations.</p>

Analysis of Customer Care Calls
• Developed a robust Speech-to-Text (STT) conversion module using Whisper AI for high-accuracy transcription,
integrating advanced Natural Language Processing (NLP) techniques to uncover customer interaction insights.
• Applied sentiment analysis and emotion detection algorithms alongside frequency analysis to classify
customer sentiments and key customer grievances and provided strategic recommendations.