Designed and deployed a serverless, cloud-native Bucket List Tracker that enables users to securely sign up, log in, and manage personal goals stored in the cloud.

Implemented authentication, a GraphQL API, and scalable data storage using managed AWS services, with optional file uploads per goal. The application demonstrates end-to-end SaaS architecture, from frontend hosting to backend APIs and persistent storage, without managing servers.

Designed and deployed a serverless event announcement system on AWS that enables users to subscribe to event notifications via email and create new events through a web interface.

The solution uses Amazon S3 for static website hosting and event data storage, API Gateway for RESTful endpoints, AWS Lambda for backend processing, and Amazon SNS for managing subscriptions and sending notifications. The architecture is fully serverless, event-driven, and built using managed AWS services without provisioning or managing servers.

Cloud Security • AWS Architecture

Intentionally designed and deployed an insecure e-commerce application on AWS to simulate real-world cloud security failures, then rebuilt the system following the AWS Well-Architected Framework (Security Pillar).

This project demonstrates secure network isolation, least-privilege access control, defense-in-depth, and application-layer protection using AWS-native services.

Designed and deployed a highly available medical web application on AWS, re-architected from a single-instance setup into a fault-tolerant, multi-AZ system aligned with the Reliability Pillar of the AWS Well-Architected Framework.

The solution uses Auto Scaling EC2 instances behind an Application Load Balancer, Multi-AZ Amazon RDS for database resilience, Route 53 health checks for DNS failover, and shared storage using Amazon EFS. The architecture is self-healing, scalable, and designed to withstand instance or Availability Zone failures.

Designed and built an event-driven AWS ChatOps system that delivers real-time operational alerts to Slack and enables secure, auditable automation without requiring AWS Console access.

The architecture integrates CloudWatch alarms with Slack via SNS, EventBridge, and AWS Chatbot, and allows permission-controlled operational actions using AWS Lambda. A read-only dashboard hosted on S3 and distributed via CloudFront provides centralized visibility, with all infrastructure provisioned using CloudFormation.

Designed and implemented an automated AWS cost optimization system aligned with the AWS Well-Architected Cost Optimization Pillar. The project simulates real-world cloud waste and applies governed, approval-based automation to detect and eliminate unused or oversized resources.

Implemented multi-tier cost detection using CloudWatch, Cost Explorer, and Trusted Advisor, followed by SSM Automation workflows with human approval gates, IAM role separation, and SNS notifications to safely clean up EC2, Elastic IPs, and storage resources.

Data Science • Machine Learning • Research

Conducted an in-depth academic study on tabular data augmentation methods for employee attrition prediction, analyzing both classical oversampling and generative approaches.

The project evaluates SMOTE, GAN-based models, and Wasserstein Conditional GAN with Gradient Penalty (WCGAN-GP), concluding that WCGAN-GP best preserves feature correlations, statistical distributions, and data privacy. A hybrid SMOTE–WCGAN-GP approach is proposed to address data sparsity and class imbalance.

Coordinated AI copilots that schedule sales meetings, balance workloads, and surface risk alerts for distributed teams.

Integrated LLM prompts, task orchestration, and guardrails to ensure accurate recommendations across CRM and analytics data.