Projects. The problem, what I built, and what changed.

Most of these started because something at work was slow, manual, or impossible to see. The numbers are real and come from production.

Southern California Edison, 2025. Shipped.

Self-service GitOps platform

GitOps for a 16-service migration. Merge-to-deploy time dropped 45%.

16: services migrated
45%: faster merge-to-deploy
35 min: build queue removed

The problem. Sixteen services were moving to Kubernetes, and every build waited in one shared 35-minute queue.

What I built. A self-service platform: Jenkins builds, ArgoCD deploys, on Kubernetes provisioned with Ansible.

What changed. Merge-to-deploy time dropped 45%, and the shared build queue is gone.

Built with Kubernetes, ArgoCD, Jenkins, Ansible, Docker.

Side project, 2025. In progress.

FreshAI

Tracks food spoilage from a photo with an 87%-accurate on-device model.

87%: holdout accuracy
16,000+: training images
$0.15: inference per month

The problem. People forget what's in the fridge and throw out food that was fine, or eat food that wasn't.

What I built. An iOS app that runs OCR and an EfficientNet-B0 classifier on-device with Core ML, trained on 16,000+ food images. AWS infrastructure is provisioned with Terraform.

What changed. 87% holdout accuracy, and moving inference on-device cut its cost from $30 to $0.15 a month.

Built with Swift, Python, YOLOv8, Core ML, Supabase, AWS, Terraform, GitHub Actions.

GitHub

UC Irvine, 2023–2025. Shipped.

ZotBins data pipeline

Ingests telemetry from 12 smart bins at about 2-second latency.

~2 sec: ingest latency
12: smart bins
$200: saved per month

The problem. Smart bins on campus send telemetry, and the batch jobs processing it ran around the clock whether there was work or not.

What I built. An ingestion pipeline from AWS IoT Core through Lambda into TimescaleDB, and 12 always-on batch jobs moved to serverless Lambda.

What changed. About 2-second ingest latency, and $200 a month less idle compute.

Built with AWS IoT Core, AWS Lambda, TimescaleDB.

zotbins.org GitHub

my home?, 2026. In progress.

Homelab setup

Segmented my home network into isolated 4-VLANs for better security and management.

4: VLANs configured
2: Managed switches deployed
1: Firewall set up

The problem. My home network was flat and stuck on my external network provider's policies, making it hard to isolate devices and enforce my own security policies.

What I built. Rebuilt my home network with a segmented 4-VLAN architecture + managed switches and a firewall, improving isolation and security for all connected devices.

What changed. Improved network security and easier management of home devices.

Built with UniFi, pfSense, Cloudflare Tunnel, Proxmox, k3s, Terraform, Ansible, Go, Linux.

GitHub