Prashant Gaur

Machine Learning & Generative AI NPS Score Prediction and Analysis Objective: Develop a machine learning-driven PowerBI dashboard to predict Net Promoter Scores (NPS) by forecasting the number of Promoters, Detractors, and Passives based on operational data from various stages (Buy, Implement, and Support). The project aims to identify key factors influencing customer satisfaction and recommend actions to improve NPS. GenAI Integration • LLM-Based Narrative Generation: Integrated GPT-4 and Anthropic Claude models to generate personalized, account-level narratives for CSM and leadership teams. • Next Best Action (NBA): Built LLM-powered NBA recommendations using Retrieval-Augmented Generation (RAG) for personalized, data-driven insights across customer segments. • Multi-Agent AI Systems: Developed multi-agent collaboration flows using both AWS Bedrock Agents and open-source frameworks (LangChain, LangGraph) to dynamically refine NBA strategies, improve decision automation, and enhance customer engagement. • Custom GPT & MCP Integration: Implemented MCP-compatible servers for interoperability with clients like Claude and ChatGPT and built custom GPT integrations backed by AWS services for secure, scalable GenAI execution. • Agent Runtime & Deployment: Deployed multi-agent orchestration flows on AWS AgentCore enabling secure runtime, authentication, and gateway-based execution of agentic workloads. Techniques & Tools: • ML: Classification models, Regression models, AWS SageMaker, Lambda, Step Functions • GenAI: AWS Bedrock, LangChain, LangGraph, RAG, Amazon OpenSearch • Evaluation: RAGAS, LLM-as-a-Judge • Observability: LangSmith, Langfuse • Engineering: Python, AWS Prompt Management (versioning), API Gateway, Lambda, OpenSearch, DynamoDB, S3 • DevOps: Docker, AWS CodeBuild, ECR, Fargate, Amplify, CI/CD

Machine Learning House Fire Risk Category Classification Project: •Objective: Develop a supervised machine learning model to classify houses into different risk categories for fire incidents. •Techniques: Utilized classification algorithms & techniques (e.g., logistic regression, random forest, XGBoost) for modeling, feature selection, feature scaling, ensemble learning, hyperparameter tuning, regularization and cross-validation techniques. House Price Prediction Project: •Objective: Develop a supervised machine learning model to predict house prices accurately. •Techniques: Utilized regression algorithms & techniques (e.g., linear regression, random forest, xgboost) for modeling, feature selection, feature scaling, ensemble learning, hyperparameter tuning, regularization and cross-validation techniques. CNN Image Segmentation Project for House Roof Prediction: •Objective: Utilized pre-trained convolutional neural network (CNN) models to develop an image segmentation model to predict house roof boundaries from satellite view images. • Techniques: Utilized image segmentation algorithms, such as U-Net, Mask R-CNN, or semantic segmentation. Image Classification & Object detection Project for House Roof Material, Roof Type Prediction and Roof Defects: •Objective: Develop an image analytics system to classify house roof images into different material categories (e.g., shingles, tiles, metal) and predict the roof type (e.g., gable, hip, flat) and also predict the defects over roof (rust, ponding, missing shingle). • Classification Techniques: Leveraged pre-trained models, such as EfficientNet , VGG, ResNet, or Inception, and fine-tuned them on the roof image dataset to benefit from learned features and speed up the training process. •Object Detection Techniques: Explored various object detection models like Faster R-CNN & YOLO to select the most suitable architecture based on detection accuracy and speed requirements.

AWS|DataEngineer|DataScientist Projects- i. Developed multiple AWS resources as per client requirements. ii. Developed multiple ETL pipeline using python. iii. Developed multiple ETL pipeline using AWS services like (S3, Glue, Lambda, SNS, MySQL) iv. Developed classification ML model for health insurance segment. DE Automated ETL Pipeline in AWS for Different Types of Finance Sources: •Objective: Develop an automated Extract, Transform, Load (ETL) pipeline in AWS to process and integrate raw data from various finance sources, enabling streamlined data ingestion, transformation and loading into database for financial analysis and reporting. •AWS Services: Lambda, S3, Glue, RDMS, Athena, SNS etc. MACHINE LEARNING Loan and Credit Card Approval Classification Project: •Objective: Develop a machine learning model to classify loan and credit card applications as approved or rejected based on applicant information and financial attributes. •Techniques: Utilized classification algorithms (e.g., logistic regression, random forest, XGBoost) for modeling.