Skip to content
albert.gorgori
Work
2025 — presentFounder, full-stack engineer

AthlioTrain

Endurance training platform for triathletes.

A training platform for triathletes, built and run solo. An AI coaching layer reads athlete profile and activity history to reason about training, with Supabase and PostgreSQL underneath.

Next.jsReactTypeScriptSupabasePostgreSQLAI SDKTailwind CSSshadcn/ui
Live siteNo public repository

Overview

AthlioTrain came out of endurance sport rather than a product brief: triathlon training generates a lot of structured data and very little useful interpretation of it.

It is an independent product, designed, built and operated by one person — which makes it the place where architecture decisions have the most direct consequences.

Problem

Triathletes accumulate workout data across three disciplines and several devices, then have no coherent view of whether the training is working.

Generic AI chat over that data is not coaching: an answer is only useful if it is grounded in the athlete's own history and profile.

Solution

  • Model training data, athlete profile and activity history as first-class records in PostgreSQL via Supabase.
  • Put an AI coaching layer on top that reads those records through defined capabilities rather than free-form database access.

AI architecture

  1. Athlete

    Athlete

  2. Agent

    AI Coach Agent

  3. Capabilities

    Training dataAthlete profileActivity analysisWorkout toolsHistorical data

  4. Data

    Supabase / PostgreSQL
Structure supplied by Albert. The coaching agent reads athlete context through defined capabilities; Supabase is the system of record.

Engineering decisions