SkinScan: what it actually takes to ship an AI-powered iOS app
Shipping an AI product to real users is harder than training the model. Here is everything I ran into building SkinScan, from camera capture to HIPAA paper trails.
~/blog
Projects, architecture, and lessons learned.
Shipping an AI product to real users is harder than training the model. Here is everything I ran into building SkinScan, from camera capture to HIPAA paper trails.
I spent six months building a TypeScript security platform. Here's what the architecture looks like, why I chose what I chose, and the mistakes I made along the way.
I got tired of writing JSON Schema by hand every time I built an MCP server. So I wrote a Go framework where your struct tags literally are your schema. Zero reflection at runtime. Here's how it works.
End-to-end encrypted email is a solved problem academically. Making it usable? That's the hard part. Here's what I learned building KrypticMail2.
A Flutter localization string extractor. Originally TypeScript (30s scan time). Rewrote in Rust (200ms). Here's the before/after, the lessons, and whether it was worth it.
I built a TypeScript framework for decentralized apps. The interesting part? The architecture is useful even if you don't care about blockchain. Here's why.
I built an expense tracker in Python that uses machine learning to categorize transactions. It's surprisingly accurate (94%) and surprisingly useful. Here's the architecture.
A commissioned enterprise CRM with real-time collaboration, offline-first mobile, and a scope that grew 3x. Here is what I learned about client projects and the architecture decisions I would change.
Applying modern object detection to dental radiographs is harder than it looks. Here is the dataset challenge, the model architecture choices, and what the deployment pipeline looks like.
Kryptic CLI is a Go-based command-line tool that streamlines common development workflows. A look at the architecture and design philosophy.
UPI is India's real-time payment backbone. I built a client from scratch in Kotlin to understand how it actually works under the hood. Here is what I found.
SRS CRM started as a simple lead management system and grew into a multi-channel outreach automation platform with AI scoring. Here is what I learned about scaling infrastructure along the way.
A custom Android launcher built with Kotlin, featuring gesture navigation, app hiding, and a minimal design philosophy.