Products & engineering
Selected work.
Products I've built, systems I've taken through delivery, and tools that have become part of my daily work.
KiNDD
Co-founder & founding engineer · 501(c)(3) nonprofit
A care-navigation product for families seeking developmental disability services in Los Angeles County. I built the web app, native iOS and Android apps, and shared backend. The web platform and iPhone app are live.
Map © Mapbox © OpenStreetMap
The product connects a family’s location to its Regional Center and nearby providers. English and Spanish support runs through the experience. The iPhone app also includes an AI assistant, next-step guidance, and shortcuts to Regional Center contact information.
My work: SwiftUI and Kotlin apps; a Python/Django/PostGIS backend for service-area matching; Bedrock agents and cited research retrieval; AWS deployment and ongoing product work. My KiNDD work includes an Anthropic grant.
Case study · Live map · iPhone app
TECHNIA
Senior Agentic AI Software Engineer · Contract
I led technical delivery of Product Structure Navigator (PSN) and ServicesIQ, owning application architecture and turning stakeholder requirements into implementation and release plans.
For PSN, my work included CI and versioned release tooling. For ServicesIQ, I built AWS infrastructure with CloudFormation, ECS Fargate, and RDS PostgreSQL, including tested-image promotion and database/file backups.
I also built an AI-powered Bitbucket PR reviewer and reusable workflows for coding agents and browser verification. That work connected AI assistance to the existing engineering and review process.
Ourself Health: Stella
Senior AI/LLM Engineer & Backend Architect
I built Stella’s AI backend with Python, Django, PostgreSQL, AWS Bedrock, and Strands agents. It supported model routing, persistent sessions, and responses streamed over SSE. The retrieval pipeline combined pgvector with semantic and BM25 search.
The surrounding systems included a HealthKit import pipeline using Celery and SQS, retrieval and response evaluations, safety checks, and Langfuse tracing with model-cost tracking. I also built a Flutter client and led the Node.js-to-Django migration and ECS Fargate delivery.
Voicebox
A macOS app I built to capture meeting audio and my microphone as separate tracks. It turns recordings into transcripts and engineering handoffs, then connects them to my Obsidian notes and coding tools. I use it nearly every day.
Built with: SwiftUI, macOS audio capture, transcription, and MCP integration.
Obsidian RAG
A local system for asking questions across my notes and following answers back to the source. Python handles retrieval, ChromaDB stores the index, and Ollama runs generation locally. MCP makes the same material available to my coding tools.
Earlier work includes Artisan Branding and 13 years of software ownership at Sherwood Development Company. Full background.