Independent Builder and AI Upskilling
Professional Sabbatical
2023 - Present- Rebuilt my personal website with Next.js, React, TypeScript, and Tailwind CSS, modernizing its architecture for performance, accessibility, and visual consistency. Benchmarked and refactored the frontend to improve mobile LCP by 0.52 seconds, introduced Partial Pre-Rendering experiments, and ensured Safari/mobile parity through dedicated proxy and CSS fallback systems. The site also showcases multiple production-grade React + TypeScript applications I built, including productivity tools, games, and utilities.
- Incorporated AI tools such as Claude Code, Cursor, Codex, and Kiro to accelerate implementation while keeping code structured, maintainable, and production-ready. I follow a spec-driven approach rather than “vibe-coding,” defining clear system requirements first, then using AI to automate repetitive work and generate high-quality code snippets. For ideation, learning, and consolidating new concepts, I regularly use ChatGPT and Gemini to refine my understanding and improve my decision-making as an engineer. This disciplined workflow enables me to build fully functional web app MVPs in under 24 hours without compromising clarity or craftsmanship.
- Invited by Google to attend Google I/O 2025, where I explored Gemini and Google’s AI ecosystem. Joined the first class of Google’s Startup School: GenAI Media training series to deepen my understanding of AI applications in creative and technical workflows. Actively participate in AI meetups and conferences both online and in person across the Bay Area, including AI-Driven Development Day, O’Reilly’s AI Codecon, Ship AI 2025 by Vercel, and the upcoming Google’s DevFest in Silicon Valley 2025 to stay current with emerging tools, trends, and expert discussions in AI development.
- Built and deployed CollectIQ from concept to production in under 2 weeks for the AWS AI Agent Global Hackathon, leading a team of three engineers to develop a serverless multi-agent AI system that authenticates and values Pokémon trading cards using AWS Bedrock (Claude Sonnet 4), Rekognition, and Step Functions. Leveraged Lambda, DynamoDB, S3, Cognito, CloudWatch, and X-Ray for scalable orchestration, storage, and observability. Designed specialized agents for OCR correction, authenticity detection, and pricing aggregation across eBay and Pokémon TCG APIs. Delivered explainable AI outputs, sub-minute analysis, and cost-efficient scalability ($0.01 per analysis at 50k runs/month) with full infrastructure-as-code via Terraform.
- Built Humanity Passport for the OpenAI Open Model Hackathon, leading a two-person team to create an AI-powered recognition platform that identifies and celebrates open-source projects contributing positively to humanity. Developed with Next.js 14, React, and TypeScript, integrating OpenAI gpt-oss-20b via Groq API, LangChain, and GitHub Octokit for repository analysis. Leveraged Prisma and Postgres for data persistence, and built a performant, responsive frontend with TailwindCSS, shadcn/ui, and Radix UI, deployed on Vercel. Delivered sub-10-second AI evaluations, dynamic SVG badges, and detailed public project pages that showcase AI-driven reasoning for each recognized repository.
- Built Reflexa AI for the Google Chrome Built-in AI Challenge 2025, a wellness-centered Chrome extension that transforms everyday reading into calm, reflective micro-sessions powered by Gemini Nano. Engineered with TypeScript, React, and Vite, and tested with Vitest, integrating all 7 Chrome Built-in AI APIs for summarization, rewriting, translation, and tone guidance with complete on-device privacy.
Next.js
React
TypeScript
TailwindCSS
Vite
shadcn/ui
Radix UI
Vitest
Prisma
Postgres
AWS Bedrock
LLM
AWS Lambda
Step Functions
Rekognition
DynamoDB
S3
Cognito
Terraform
Vercel
Gemini Nano
Groq API
LangChain
GitHub Octokit
Chrome Built-in AI APIs
AI Agents
Multi-Agent Systems
Generative AI
Context Engineering
Spec-Driven Development
Performance Optimization
Accessibility
Serverless Architecture
Hackathons
Open Source