Work
Production systems I designed and built at AiDash, and the occasional side project.
§1 At AiDash
2025 – nowCRIS 2.0CRIS shows power utilities where wildfire and weather risk is highest on their network. This is the service that feeds it. I have worked on CRIS since its inception and owned its backend architecture as it grew from 1 to 50+ enterprise tenants.Tenants: 1 → 50+
2025 – nowWMP CopilotA RAG pipeline that lets utility companies ask questions of their own wildfire mitigation plans. I designed and deployed it for 100+ tenants, across a Java and Python microservices architecture.Tenants: 100+
2023 – 2025Serverless data archivalTwo tables were taking in about 750K records every 6 hours, and the data stopped mattering after three days. A scheduled, per-client pipeline on Lambda, SQS and Fargate now moves everything older than 14 days to S3 as Parquet. Primary RDS storage went from 400 GB to 3 GB.RDS storage: 400 GB → 3 GB
2023 – 2025Multi-tenant ingestion on LambdaCRIS 1.0 ingested each client's data through its own Lambda. I replaced that with one multi-tenant system: a single codebase, 300k+ records every 6 hours inside Lambda's limits, and onboarding that needs almost no engineering work.Volume: 300k+ records / 6 hours
§2 Side projects
2025PlaylistGenie: AI-Powered Playlists from Just One PromptHow I built an AI Spotify wrapper in a weekend that turns "late-night road trip vibes" into the perfect playlist in secondsReact, Python FastAPI, Spotify API