My Projects
AI-Powered Support Chatbot with AWS Bedrock and Claude
Architected and guided the implementation of an advanced AI chatbot leveraging AWS Bedrock and Claude to provide intelligent responses based on the application's support articles, significantly enhancing customer support efficiency.
Key Achievements:
- Designed serverless architecture using AWS Bedrock, Agents, and Knowledge Bases for an efficient chatbot
- Implemented vector database with AWS OpenSearch for optimized support article retrieval
- Developed ETL pipeline to process support articles from PostgreSQL to S3 via JSON transformation
- Integrated Claude AI model via AWS Bedrock for context-aware and accurate responses
- Improved response times and accuracy, enhancing overall user satisfaction
Technologies Used:
B2B SaaS Platform for Hospitality Industry
Co-founded and developed a B2B SaaS solution that became a market leader in providing real-time market data for the hospitality sector.
Key Achievements:
- Rapidly grew customer base to over 3,000 within the first month of launch
- Architected a scalable backend using AWS, Ruby on Rails, and PostgreSQL
- Developed innovative algorithms for real-time data analysis and reporting
- Successfully positioned the company for acquisition through consistent growth and market leadership
Technologies Used:
Data Centralization and Analytics Dashboard
Designed and implemented a comprehensive dashboard application to centralize data from multiple sources and improve decision-making processes.
Key Achievements:
- Developed an efficient web-crawler for automated data collection from various sources
- Implemented RESTful APIs to integrate data with internal tools and Excel
- Migrated from static Excel-based reporting to a dynamic, cloud-native application
- Significantly improved data consistency and accessibility across the organization
Technologies Used:
Enterprise Application Serverless Migration
Led the migration of a legacy application to a complete serverless architecture on AWS, significantly improving performance and reducing costs.
Key Achievements:
- Migrated monolithic API to a GraphQL-based microservices architecture on AWS Lambda
- Implemented a modern React-based frontend hosted on AWS S3 and CloudFront
- Transitioned from traditional RDS to a serverless database solution
- Achieved 50% reduction in monthly infrastructure costs while improving application performance
Technologies Used:
Monster Research Incorporated
Monster Research Incorporated is a Next.js application that generates, displays, and ranks unique monsters using AI.
Key Achievements:
- AI-powered monster generation using OpenAI's GPT and DALL-E
- Real-time leaderboard and monster carousel
- User authentication with Clerk
Technologies Used:
Multi-Tenant Serverless Database Architecture
Architected and implemented a shared serverless database cluster to support multiple applications across different cloud accounts, optimizing resource utilization and cost efficiency.
Key Achievements:
- Designed a scalable multi-tenant architecture supporting both production and non-production environments
- Consolidated multiple database instances into a single, efficient cluster
- Optimized high-traffic queries for improved performance across all tenants
- Achieved significant cost savings and performance improvements through resource sharing and optimization
Technologies Used:
Personal Portfolio Website
Designed and developed a modern, responsive portfolio website to showcase professional experience and projects.
Key Achievements:
- Implemented a sleek, user-friendly design using Next.js and Tailwind CSS
- Integrated dynamic content management for easy updates and maintenance
- Optimized performance and accessibility to ensure a great user experience across devices
- Implemented serverless functions for backend operations like contact form submission
Technologies Used:
PhotoMuse: AI-Powered Image Search and Analysis
PhotoMuse is an advanced image search and analysis application that uses AI to generate descriptions and tags for images, and then allows for similarity-based searching using natural language queries.
Key Achievements:
- Participated in Microsoft RAG-Hack 2024, developing an innovative image search application
- Implemented image upload and automatic description generation using Azure Computer Vision and GPT-4
- Integrated automatic tagging and vector embeddings for efficient similarity search
- Developed natural language querying and AI-refined search capabilities
- Provided confidence scoring and explanations for search results to enhance user experience
Technologies Used:
Mary Woodward PSO Dynamic Calendar Application
Developed a modern, serverless calendar application for Mary Woodward Parent Support Organization (PSO) that transforms the school's public Google Calendar into an interactive, user-friendly interface for parents and staff.
Key Achievements:
- Architected and implemented a serverless application using SST (Serverless Stack) and AWS services to fetch and process iCal data from the PSO's public Google Calendar
- Created a responsive, dynamic web interface that displays up-to-date school events in an easily digestible format
- Implemented a feature allowing users to download a PDF version of the calendar for offline access
- Designed the application to be embeddable within the existing PSO website, ensuring seamless integration