Here's what I've been building.
AI-powered business card scanner that turns a photo into importable contact data. Uses Claude's vision model with BAML for structured field extraction, then exports as a vCard (.vcf) for Apple Contacts or a CSV for Google Contacts. Implements a bring-your-own-key design that never stores API credentials in plaintext, pairing practical utility with privacy-first architecture.
AI-powered advice service delivering thoughtful guidance on interpersonal challenges. Built with a three-stage LLM pipeline: input screening, response generation, and voice-matching refinement. Demonstrates practical multi-agent orchestration and sophisticated prompt engineering for consistent, empathetic output.
AI-powered receipt parser transforming photos into structured spreadsheet data. Combines computer vision and LLMs to extract line items, totals, and merchant details with direct Google Sheets integration. Eliminates manual expense tracking through automated data extraction and formatting.
Intelligent recipe assistant built with retrieval-augmented generation (RAG). Searches a vector database of curated recipes using semantic similarity, falling back to LLM generation when needed. Demonstrates practical implementation of hybrid AI systems balancing accuracy with flexibility.
Opinionated filesystem architecture designed for AI agent development. Organizes work into categories (features, bugs, refactors, tests) with executable instruction documents that AI agents can interpret and run directly. Includes GitHub Issues synchronization and workflow automation for AI-assisted development.
AWS Certified Solutions Architect (2025)