• Designed and built Quinsight, an AWS-based NL→SQL analytics platform for QuinStreet, enabling business users to query Amazon Redshift data in natural language using AWS Bedrock (Claude 3.5) in a production-grade environment.

• Implemented a request router and identity-aware inline agents with Bedrock AgentCore and AWS Strands to route traffic between FAQ fast-path and NL→SQL workflows, enforcing tenant/role-based policies and least-privilege tool access.

• Built the FAQ fast-path service on DynamoDB with parameterized SQL templates and Redshift Data API integration, delivering low-latency responses for recurring questions and reducing load on the NL→SQL pipeline.

• Configured enterprise-grade authentication and authorization for Quinsight using Azure Active Directory, implementing OAuth2/OIDC flows, role-based access control (RBAC), and secure token validation layers to ensure protected access for internal and partner applications.

• Integrated full-stack observability using OpenTelemetry, OpenLLMetry, Langfuse, and GenAI observability tooling, enabling trace-level visibility across NL→SQL agents, Lambda workflows, Bedrock calls, and Redshift queries; improved debugging efficiency and reduced MTTR for production issues.

• Contributed to infrastructure-as-code and deployment automation with AWS CDK/Terraform for API Gateway, Lambda, S3, DynamoDB, and Redshift integrations, ensuring reproducible environments across dev/stage/prod.

• Partnered with frontend teams to expose a single automation-friendly router API, where React calls one endpoint and the backend orchestrates all downstream agents, tools, and visualisation workflows.

• Developed an AI-driven IT ticket automation system used by 1,000+ university staff and students, reducing issue resolution time by 65% through intelligent ticket classification, real-time status tracking, and smart routing using LLM APIs via Amazon Bedrock, achieving over 90% classification accuracy.

• Leveraged AWS cloud services (EC2, Lambda, Redshift, SageMaker, RDS) to deploy predictive analytics solutions for student performance tracking and enrollment forecasting, cutting infrastructure costs by 25% via efficient cloud resource orchestration.

• Adopted Agile/Scrum and full SDLC methodologies, managing sprints using Jira, implementing CI/CD pipelines in Jenkins, and maintaining code repositories in Git/GitHub, ensuring continuous delivery and streamlined cross-functional collaboration.

• Created secure administrative dashboards using React, OAuth2, and Chart.js, integrated with AWS CloudWatch for real-time system monitoring, performance metrics, and automated anomaly detection, improving operational visibility and decision-making.

• Built LLM-powered RAG platforms for academic content retrieval, research paper summarization, lab documentation synthesis, and compliance reporting, reducing faculty and research staff review time by 35% while supporting human-in-the-loop validation for accuracy and governance.

• Designed and deployed GenAI Agents for automated technical report generation, equipment documentation triage, and knowledge-base indexing using GPT-based and Gemini-based models, improving workflow efficiency for academic departments and engineering labs.

• Developed predictive analytics models (Logistic Regression, XGBoost) for student performance forecasting, research resource allocation, and equipment-failure risk scoring, increasing predictive accuracy by 18% and enabling proactive decision-making for academic and technical operations.

• Led the design and implementation of a customer churn prediction model using XGBoost, reducing client churn rates by 15% and improving customer retention and satisfaction across IT consulting engagements.

• Developed and deployed Natural Language Processing (NLP) solutions for sentiment analysis of client feedback and service tickets, improving customer feedback analysis efficiency by 20% and strengthening client relationship management.

• Built and optimized deep learning models including Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) for image recognition and sequential data analysis, enhancing predictive accuracy and model performance by 30%.

• Collaborated with cross-functional delivery teams (data engineers, consultants, product managers) to integrate AI/ML models into enterprise IT solutions, driving a 30% increase in user engagement for client-facing applications.

• Performed data preprocessing, feature engineering, and model optimization to achieve more than 90% accuracy rates across multiple AI/ML use cases, including predictive analytics, recommendation engines, and anomaly detection.

• Evaluated and contributed to the selection of ML frameworks, influencing the enterprise adoption of TensorFlow for scalable deep learning projects within Hexaware’s IT consulting portfolio.

• Applied full-stack deployment skills (HTML, CSS, JavaScript, Flask API) to integrate ML models into client web applications, ensuring seamless end-to-end deployment and usability.

• Deployed scalable AI/ML solutions on AWS Cloud, leveraging SageMaker, EC2, and S3 to support real-time inference, model training, and data storage for enterprise clients across banking, insurance, and healthcare domains.