Product-minded software engineer from India. 9+ years in Ruby on Rails. I notice friction, research the problem, and build tools that remove it. Every project below started with a real frustration, not a side-project idea list.
I kept hitting free-tier limits on error monitoring SaaS, and realized a gem living inside the app has access to things external services never will. So I built a self-hosted alternative that detects silently rescued exceptions, captures instance variables at the crash site, and snapshots GC/YJIT/connection pool health at the moment of failure. Free forever, no vendor lock-in.
13K+ downloads, 3,000+ tests, 59 releases, 5 community contributors. Live Demo · Docs
Most AI coding tools write code for you, which feels productive but builds no lasting skill. I studied the research (Bloom, Vygotsky, Oakley, Ericsson) and built a Claude Code tutor that makes learners write code while the AI asks the right questions. Spaced repetition keeps concepts from fading, adaptive difficulty keeps challenge in the sweet spot. 17 skills, 3 AI agents, 15 knowledge bases.
Most Claude Code starter kits are over-engineered. I studied what actually works (ETH Zurich research on context files, Arize's rule optimization study) and built something minimal. Drop it into any project, it auto-detects your stack, loads the right rules, and gives you 15 purpose-built skills. Includes a context audit skill that shows you exactly what is eating your context window.
Setting up a new Mac for development takes most people a full day. This does it in one command: 325 packages, 17 apps themed consistently, work/personal identity split. 170+ tests, fully idempotent.
Zero-config Zed editor setup. Run zs in any project, it detects the stack, installs debug dependencies, creates task configs, and sets up git hooks. Saves about 15 minutes per project.