My Project Portfolio

Here are some of my most representative projects. Click any card to view the full story, architecture, and my contribution.

2026-02

Twinkle Voice: Scenario-Based Speech Training with Real-Time Feedback

Feb 2026 – Present

A mobile-first speech training product that combines scenario missions, turn-based voice interaction, and actionable pronunciation/prosody feedback.

PythonFastAPIASRProsody AnalysisPronunciation ScoringPrompt Engineering+1

2025-08

Real Talka: Enterprise AI Voice Sales Assistant

Award

Watch Demo

Real Talka: Enterprise AI Voice Sales Assistant

Aug 2025 – Present

Taipei Youth Bureau 2025 Public Policy Award

A multi-tenant B2B SaaS platform that brings AI agents into voice-based sales calls, combining speech, knowledge management, and omnichannel delivery.

PythonFastAPIPostgreSQLWebRTCSIPLLM API+3

2025-05

Watch Demo

Voice bear: Hybrid On-Device Speech Recognition

May 2025 – Jul 2025

Built a CoreML encoder + GGML decoder pipeline for the Breeze-ASR-25 model to deliver low-latency, Taiwan-accent optimized speech recognition on Apple Silicon.

PythonCoreMLGGMLMetalApple Neural EngineWhisper.cpp+2

2024

Lumacool Delivery: Self-Managed Food Delivery Ecosystem

Independently developed food delivery platform to solve high commission fees and data opacity issues from mainstream platforms, empowering merchants with cost reduction and marketing autonomy.

PythonFastAPIReactNext.jsTypeScriptPydantic+5

2023

Watch Demo

Meowda: AI-Powered Houtong Cat Village Map Navigation and Story Interaction System

Led a team to develop a LINE chatbot that uses AI to identify specific cats in Houtong, combining storytelling with gamified mapping to enhance visitor interaction.

PythonFlaskFastAPIYOLOv7DockerGoogle Cloud Run+3