Vyom

Project Overview

Vyom is a feature-packed wellness app that brings yoga posture tracking, guided yoga tutorials, and smart meal planning into one unified experience. It runs real-time posture correction using on-device machine learning optimized for Arm hardware, ensuring fast, private, and reliable performance. The app also includes seamless Android TV integration, making guided yoga sessions more immersive on larger screens.

Vibe Coding

Used Kiro for rapid prototyping. Workflow: describe → generate → test → refine. Helped with:

• Android TV casting
• ML-based nutrient tracking
• Handling multiple authentication flows Best output: initial Android TV casting architecture via platform channels.

Agent Hooks

Built hooks to automate repetitive tasks. Main automations:

• Server formatting/style checks
• Database migration + model generation pipeline Result: fewer mistakes, consistent backend, faster iteration.

Spec-Driven Development

Wrote short specs before big features. Specs included purpose, I/O, endpoints, edge cases. Used for: Android TV casting ML nutrient tracking Result: cleaner, predictable code vs open-ended chat.

Steering Docs

Used rolling steering docs (rebuild after each major update). Removed old docs to avoid outdated context. Gave Kiro a fresh, accurate view of the project. Result: better edits, fewer misunderstandings.

Built With

• flutter
• node.js
• postgresql