AI 육아 · StartupFlagship

Parentlyze

An AI parenting service that turns a single parent-child conversation into a tailored parenting report — the iOS/Android app is the product, backed by the web/Admin/backend and an in-house STT/AI engine. Co-founded 2025-06, CTO.

Team · 3Duration 12 moContribution 85%2025-062026-04
Key Metric: Co-founder CTO · owns the app, the web, and the in-house STT/AI
User Scale: In beta · iOS/Android app first
Engineering Role
FullstackMobileAI/MLInfraEnd-to-End
AI 육아 · Startup
Parentlyze
AI
#Mobile#Web#Backend#AI

Overview

In June 2025 I co-founded Potentivo Lab and build this AI parenting startup as CTO. Record a parent-child conversation and it is diarized, transcribed, and turned by multiple LLMs into a tailored parenting report and coaching missions. The real consumer product is the Flutter iOS/Android app, backed by a Next.js web layer (the app's WebView content, the AI report backend, the operations Admin, and the landing) and an in-house GPU STT/AI engine — all three surfaces share one Supabase backend.

Product Surfaces

Mobile app (Flutter · iOS/Android)

Flutter · Dart · WebView hybrid · WorkManager

The face of the actual product users use.

  • Auth/recording/payments/push are native Flutter; home/missions/reports are WebView-embedded
  • Background long-form audio uploads via a WorkManager + SQLite local queue
  • iOS/Android IAP receipt verification · OneSignal push · deep links
  • Shipped on the App Store and Google Play · in production

Web · Admin · Backend (Next.js)

Next.js 16 · Supabase RLS · multi-LLM · Toss

App WebView content + AI report backend + operations Admin + landing, in one codebase.

  • Embeds the app's WebView screens (home/missions/reports) via a token bridge
  • Multi-LLM (OpenAI·Claude·Gemini·Cohere) report/mission-generation backend + STT trigger · payment webhooks
  • Operations Admin — recordings/STT · reports · prompts · users · payments · discounts · notifications · growth CRM
  • Supabase RLS multi-tenancy + Public·WebView·Admin auth tiers

STT · AI engine (FastAPI · GPU)

FastAPI · faster-whisper · Qwen3-ASR · pyannote · CUDA

An in-house AI server that transcribes, diarizes, and acoustically analyzes conversation audio.

  • faster-whisper (large-v3) + Qwen3-ASR in parallel, with pyannote diarization
  • asyncio + per-engine semaphores to avoid CUDA OOM
  • Async job queue → Supabase callback, dual v1/v2 APIs
  • A self-hosted STT pipeline with zero OpenAI dependency

Highlights

  • Co-Founder & CTO · founded 2025-06, owning product and tech
  • App (Flutter) · web/Admin/backend (Next.js) · in-house STT/AI (FastAPI GPU) — three surfaces on one Supabase
  • Record → diarized transcription → multi-LLM tailored parenting reports and coaching missions
  • Shipped on the App Store and Google Play · Toss / IAP payments

Timeline

2025-062026-04
Role: Co-Founder / CTO — owning product and tech across app, web, backend, and AI
Contribution: 85%
Collaboration: Team · 3

Tech Stack

</>Flutter app · Next.js web/Admin/backend · in-house STT (FastAPI GPU) · Supabase

FlutterDartNext.js 16React 19TypeScriptSupabasePostgreSQL (RLS)FastAPIfaster-whisper (large-v3)Qwen3-ASRpyannote.audioCUDAOpenAI SDKAnthropic SDKGoogle Gemini SDKToss PaymentsOneSignal

Links