Adonis AI — Hyunjun Kim

Adonis AI

A camera-based AI personal trainer (hackathon prep) — pivoted to Money Navigation at the event

Experiment

2026.10 / Solo (AI as the team) / iOS · Android

Problem: People working out alone have no one to check their form.
My role: Planning · setting the bar · directing AI to build
Stack: Flutter · Google ML Kit (Pose Detection) · Supabase

Results

ML Kit pose detection (no server upload): On-device; ML Kit pose detection (no server upload)
Entered and finished U-Hackathon 2026 solo: Hackathon; Entered and finished U-Hackathon 2026 solo

Idea / Problem

People working out alone in a Bangkok condo have no one to check their form. I conceived an AI personal trainer that reads posture from the phone camera alone, built as prep for U-Hackathon 2026 in Bangkok (Oct 15–16, 2026).

The Plan & Bar I Set + Directing AI

Bar: the workout video never leaves the device (latency, privacy). So with AI I chose on-device ML Kit pose detection over server analysis, and directed the app skeleton plus the camera–ML Kit wiring.
Being a short hackathon, I decided what not to build first and scoped instructions to one core thing: pose detection.

Results

Integrated on-device pose detection via ML Kit
Entered and finished U-Hackathon 2026 solo (the final on-site submission pivoted to a traveler payment/spend assistant — that finished project is Money Navigation)

Retrospective

In a short window, narrowing the scope of instructions decided the outcome. Focusing on one core thing helped me scope later projects too.

Note: The code itself is early-stage; this is a supporting example of AI/ML use and finishing a hackathon solo.