What I built, key architecture choices, and source you can inspect.
AI agents · Desktop & backend
Coretex
Pre-1.0 · Source available
A local-first workspace bringing AI agents, projects, communication, personal analytics, and developer operations into a Windows desktop application.
My contribution
Built and maintain the desktop/web workspace, shared React application layer, and TypeScript Brain service.
Architecture choices
- Share the application layer across Electron and Next.js clients.
- Use PostgreSQL and Prisma in the Brain service, with Docker-backed local development.
- Keep credentials and local application data outside the public repository.
TypeScript · React · Electron · PostgreSQL · Docker
AI integration · Safety contracts
Core AI
Public TypeScript package
A small provider-neutral package for conversation safety and metered-usage authorization, used at the boundary between applications and private AI services.
My contribution
Built and maintain reusable input-isolation and authorization contracts for chat, voice, and embedding requests.
Architecture choices
- Isolate user-supplied text before it enters a prompt data field.
- Require a typed usage-authorization decision before metered operations.
- Keep provider SDKs, production policy, prompts, and customer data outside this package.
TypeScript · ES modules · Input isolation · Usage authorization
Full-stack · APIs & data
Health Tracker
Source & local setup available
A personal health tracking application with a Next.js interface and a Spring Boot backend for workouts, nutrition, and other tracking workflows.
My contribution
Built the full-stack application, connecting tracking interfaces to Java backend services and persistent storage.
Architecture choices
- Separate the Next.js frontend from the Spring Boot API.
- Use PostgreSQL for relational storage and LocalStack for local S3-compatible development.
- Document service ports, startup steps, and deployment configuration for reviewers.
Java · Spring Boot · Next.js · PostgreSQL · Docker
Full-stack · AI & service integrations
Portfolio application
Live at mdcran.com
The application behind this website: a responsive portfolio with managed content, API integrations, streaming AI chat, voice features, and analytics.
My contribution
Designed, built, and maintain the Next.js application and its content and integration routes.
Architecture choices
- Use server-rendered pages alongside interactive React components.
- Store editable portfolio content in MongoDB.
- Integrate streaming chat and speech services through server-side API routes, keeping credentials on the server.
Next.js · React · TypeScript · MongoDB · REST APIs