Building Avya — the AI fitness app named after my son Avyaansh, powered by 14 years of real calisthenics experience and Anthropic AI.
I named my AI fitness app after my son. His name is Avyaansh. The app is called Avya.
That single decision changed how I build. When a product carries your child's name, every shortcut feels like a betrayal. Every lazy design choice feels personal. You do not ship something half-baked when your son's name is on it. You build it like it matters — because it does.
Avya is an AI fitness assistant. Not another calorie counter. Not another workout log with a chatbot bolted on. It is an intelligent system trained on 14 years of real calisthenics experience, designed to give people the kind of personalized fitness guidance that currently only exists if you hire an expensive coach or get lucky enough to know someone who actually lives this life.
The Problem Avya Solves
Fitness advice is abundant. Open YouTube and you will drown in workout videos. Search Google and every fitness website will tell you to eat protein and lift weights. There is no shortage of information.
The shortage is in personalized fitness intelligence.
Here is what I mean. A 22-year-old guy in Pune who weighs 60 kg, has never trained, works a desk job, and wants to start calisthenics — he does not need the same advice as a 35-year-old woman in Delhi who has been running for 5 years and wants to add strength training. The information they find online will be roughly the same generic advice. What they actually need is radically different.
A good coach adjusts everything — exercise selection, volume, progression, nutrition timing — based on the individual. That is what makes coaching valuable. Not the knowledge itself, but the application of knowledge to a specific person with specific constraints and specific goals.
Avya does that. It takes the knowledge I have accumulated over 14 years — not textbook knowledge, lived knowledge from training my own body through Navy service, injuries, plateaus, and breakthroughs — and applies it intelligently to whoever is asking.
Fitness information is free. Fitness intelligence is rare. Avya exists to close that gap.
The Technical Stack Behind Avya
Avya runs on a stack I chose for longevity, not trendiness.
Anthropic API (Claude) — the brain. Claude handles the conversational intelligence, the ability to understand context, remember previous interactions, and generate responses that sound like a knowledgeable coach, not a robot reading a textbook. I use Sonnet for most interactions and Opus for complex programming and deep analysis. The model quality from Anthropic is the reason Avya sounds human.
Next.js 14 — the frontend and server layer. App Router, server components by default, deployed on Vercel. Fast, reliable, and built to last. No exotic framework that will be abandoned in two years.
Supabase (PostgreSQL) — the data layer. User profiles, workout history, conversation logs, progress tracking — all in Postgres with Row Level Security so every user's data is protected at the database level, not just the application level.
Voice interface — because the best interaction with a fitness coach should feel like talking to a person, not typing into a form. Voice input makes Avya accessible while training, when your hands are on a pull-up bar, not a keyboard.
Every technology choice answers one question: will this still work in 10 years? If the answer is uncertain, I pick something else. Avya is not a weekend project. It is infrastructure that needs to outlast trends.