-
-
KAEL by NEXVYVE — an AI fitness coach for Spanish speakers, installed straight from the browser.
-
Chrome → Add to home screen → done. No store listing, no review queue, no 30% cut.
-
dailyFunnelDigest, 08:00 Europe/Zurich. The numbers are computed by code over Firestore; the reading is written by Gemini.
-
A full plan in under five minutes. The plan engine is deterministic by design — Gemini is used for judgment, not for reproducible work.
-
/api/kael. The coach answers with the user's full history in context: what they trained, what they ate, what they skipped.
-
Two scheduled Cloud Functions in europe-west1, running since 18 July 2026. No errors.
-
/api/ai/food-scan on gemini-2.5-flash. Photo in, food identified and macros computed — about $0.0012 per scan.
-
KAEL (Android and web) and NEXVYVE (iOS) share a single community. The bubbles in this feed are in five languages.
-
/api/ai/body-analysis. Gemini reads the photo, scores it and names the lagging areas — and the training program updates to match.
-
Five purchases at the public $39/year price. The gap between 23 June and 26 July is the Venezuela earthquake. Two sales since.
I've stood on stage as overall champion at the Arnold Classic Europe. These days I spend most of my hours on a construction site in Zürich. KAEL is what I build at night, after the shift is over.
My name is Bleiquel Colina. I'm an IFBB Pro athlete and a sports-nutrition specialist (I trained through Barça Universitas at FC Barcelona). In 2019 I was national champion of Venezuela and South America; in 2020 I took the overall at the Arnold Classic Europe. But the part I'm actually proud of is quieter: for five years I've coached people one by one — their training, their food, and the unglamorous stuff nobody posts about, like discipline and breaking habits that took years to build.
I left Venezuela for Spain, and a year ago I landed in Zürich. New country, new job, no gym career waiting for me here. Honestly, I started building KAEL when I came across this challenge — it was the push I needed to turn ten years of coaching into something that doesn't stop at one client at a time. I didn't want to help one person a week. I wanted to help thousands, and do it properly.
What KAEL is. An AI coach that carries my method. You photograph your body and your meals, tell it where you want to go, and it builds you a real training and nutrition plan, then adjusts it with you every day. One rule I refused to break: KAEL suggests, you decide — it proposes each change and you accept it with one tap, never behind your back. It speaks Spanish, English and German. What I'm really after is a feeling: that you have a real coach in your corner 24/7, and a community walking the same road, so nobody does it alone. Discipline, real change, better health, and the confidence that comes with all three.
Why it runs on Gemini. The Gemini API is the actual brain. It runs the daily coaching conversation, reads your body scans and meal photos, and adjusts your priorities from what it sees. The judgment calls I used to make by hand for every client — reading a body, sizing a plate, deciding what to push on — Gemini now makes live, in production, on Google Cloud.
This was never just an app for me. It's how I keep doing what I'm good at, for people who could never afford a coach like me.
What it does
You install KAEL from the browser onto your Android home screen in about ten seconds. No store, no download, no gatekeeper.
You answer a short onboarding and in under five minutes you have a training and nutrition plan — in Spanish, English or German. From then on:
- Every day it tells you the single next thing to do, based on the actual time of day and what you still have pending.
- You photograph your plate and Gemini identifies the food and computes the macros. No database lookup, no searching for "arepa" in a food list.
- You photograph yourself and Gemini analyzes your body composition, tells you which zones to prioritize, and adjusts your program accordingly. A month later it compares the two and shows you what actually changed.
- You ask the coach anything about your own plan and it answers with your full history in front of it — what you trained, what you ate, what you skipped.
- And the community is one: KAEL users on Android and NEXVYVE users on iPhone see each other, compete, and message each other directly.
How we built it
Next.js 14 (App Router) and TypeScript on Vercel, with a 3D interface on the front. Firebase for Auth and Firestore. Stripe for payments. Installed as a PWA instead of shipped through a store. Built solo, in the hours most people are asleep.
Gemini 2.5 Flash runs four endpoints in production:
/api/kael— the coaching conversation, with the user's full context/api/ai/food-scan— photo of a meal to identified food and macros/api/ai/body-analysis— photo to body composition and training priorities/api/ai/evolution-compare— progress comparison over time
And two scheduled Cloud Functions that run the business rather than the product:
nightlyHealthcheck (06:00) checks system health and decides whether to alert
me; dailyFunnelDigest (08:00) aggregates the funnel, and Gemini writes my
revenue report and decides whether it's worth waking me up.
One deliberate architectural decision: the plan engine itself is deterministic, not LLM-generated. A coach that invents a different program every time you ask it is not a coach. Gemini is used where judgment and language belong — vision, conversation, analysis, and business reporting. Rules are used where reproducibility belongs. That boundary is the most important design choice in the product.
Pre-existing code, declared: NEXVYVE, my iOS app, existed before May 19, 2026. KAEL is a separate codebase in a separate repository, first commit June 1, 2026. What carried over is pure domain logic ported by hand — the points system, the plan engine, and the exercise and food catalogs. Everything else was written during the Hackathon.
Challenges we ran into
On June 24, 2026, a catastrophic earthquake struck Venezuela, where I am from. People I knew died.
I stopped working on KAEL. I spent those weeks building Faro VE (https://faro-ve.com), a free installable web app for reporting and finding missing people — a geolocation map, a search across more than 47,000 reported people, and an offline first-aid guide with 34 protocols that works with no connectivity, because in a disaster the networks are the first thing to fail. It is connected to the International Red Cross Person Finder network through the open PFIF standard. It is still running and it can be independently verified.
That is why this project's revenue clusters in June, stops, and resumes in late July. My last sale was June 23. The earthquake was June 24. The gap is visible in the Stripe export attached to this submission, and I would rather explain it than smooth it over.
The other challenges were smaller but real: making a web app feel premium enough that nobody misses a native one; keeping Gemini costs near zero so the unit economics work at $4.99 a month; and building a social layer that works across two separate codebases on two different platforms.
Accomplishments that we're proud of
A real business, launched, with paying customers at a public price — not a prototype and not a waitlist.
AI that runs the operation, not just the product. Every morning Gemini reads my own business and decides whether the state of it is worth interrupting me for. I don't watch dashboards.
Zero app store. No 30% cut, no review queue, no gatekeeper between someone on a cheap Android in Latin America and a real coach.
Ten years of coaching one person at a time, finally encoded into something that can serve thousands without losing the method.
And Faro VE, built inside this same window, still serving people.
What we learned
Ten years of gut feeling don't transfer to a model on the first try. Teaching KAEL to coach the way I coach took far more nights than building the app did.
Distribution beats features. Every customer so far came from an audience I built over years of publishing fitness content, at $0 in paid acquisition — and I have never even run a campaign to it.
AI is worth more pointed at the operation than at the product. The feature everyone builds is a chatbot. The thing that actually changed how I work is a scheduled function that reads my business every morning and decides whether to interrupt me.
And people trust it more precisely because it asks before it acts. KAEL proposes every change and you accept it with one tap. Never behind your back.
What's next for KAEL by NEXVYVE
Switch on the distribution I already have. KAEL still has no Instagram account of its own, and I have never run a launch campaign to the 187,000 people who follow me. That is the cheapest growth available to this business and it is entirely undone.
Then: finish the Founder cohort of 100. Ship the retention agent — an AI that notices a user going quiet and intervenes before they churn, instead of after. Add Mercado Pago so people across Latin America can pay in local currency.
And the milestone that decides everything: roughly 800 paying subscribers. That is the point where I can put down the tools, leave the construction site, and do this full-time — helping people change their bodies, their health, and how they see themselves.
Log in or sign up for Devpost to join the conversation.