Senior Generative AI R&D Engineer · Audio & Music AI Developer · Independent Musician
Building practical AI systems and audio software for creators, communities, and intelligent workflows.

I build practical AI and audio software that turns research into working tools for creators. My recent work combines local ONNX inference, C++ and Flutter desktop applications, DAW integration, source separation, melody transcription, live pitch tracking, and cross-platform audio plugin development for AAX, VST3, and AU.

Alongside music technology, I develop production-facing generative AI systems—including RAG and agent workflows, structured outputs, Physics AI, and multilingual knowledge products—with an emphasis on traceable automation, cross-platform delivery, and interfaces people can use without machine-learning expertise.

Open to collaborations in generative AI, audio software, creator tools, and music technology.

Generative AI     RAG, agents, structured outputs, traceable workflows
Music AI          Source separation, transcription, pitch and tempo analysis
DAW tooling       Pro Tools PTSL, stem workflows, AAX, VST3, AU
Desktop products  Flutter, C++, ONNX Runtime, macOS and Windows packaging
Web systems       Astro, React, TypeScript, FastAPI, GitHub Pages

Project	What it does
TrackMate	An AI-assisted Pro Tools stem-import workflow that recognizes instruments from filenames and audio, standardizes track names, categories, and colors, detects BPM and tempo maps, identifies fake stereo for mono conversion, and imports and verifies everything through Avid's official PTSL—all locally.
karaoke-battle	A local-first karaoke scoring app for streamers that turns YouTube links or media files into singable tracks with vocal separation and melody transcription, displays live pitch against scrolling notes, and scores pitch, rhythm, stability, vibrato, dynamics, and sustained notes on macOS and Windows.
Semitone	A multilingual browser tool for generating chord progressions from LΛMPLIGHT’s Shasavistic music theory, with just intonation, 41edo and 12edo playback, instrument-specific performance guidance, interactive synthesis, and MIDI export.
ai-interview	A multilingual AI engineering interview platform with source-linked flashcards in Traditional Chinese, English, and Japanese, structured AI-assisted study notes, community contributions, and daily incremental updates via GitHub Actions.
PIE — Physics Informs Everything (private)	A PhysicsNeMo-powered MCP bridge and nine-skill research workflow that helps AI agents formalize physics problems, inspect models and literature, process CAD and mesh inputs, validate PhysicsNeMo implementations, and run traceable experiments.
music-detection	A privacy-first browser music analysis tool that detects BPM, time signature, key, and vocal pitch range with ONNX inference, previews beat and pitch results, and exports Tempo + Lead Vocal MIDI, with an optional local Python engine.
ktv-charts	A multilingual karaoke chart explorer that aggregates daily rankings from Taiwanese and Japanese platforms, enriches songs with previews, ISRC credits, BPM, vocal-range, and music-tag metadata, and publishes searchable history through GitHub Pages.
HT-76	A JUCE-based 1176-style FET compressor for macOS and Windows, with Rev D and experimental Rev H circuit models, VST3, AU, and AAX builds, plus switchable classic rack and real-time analysis views.
JS_Inflator (customized fork)	A cross-platform JS Inflator fork with macOS and Windows VST3, AUv2, and AAX builds, installer packaging, and audio/UI regression testing; its AAX adaptation targets Pro Tools Developer and remains in testing.
software-salary	Taiwan software salary insights based on 709 analyzed salary samples, with market percentiles, company rankings, experience-based estimates, and interview guidance.
Private RAG / Agent Systems (private)	Internal AI systems for retrieval-augmented generation, document workflows, structured outputs, and production-facing automation.

• 2026-11-13 — Updated Hikari-Tsai/warranty-form: Adjusted grid field spacing so form items aligned correctly within the layout.
• 2026-11-13 — Updated Hikari-Tsai/karaoke-battle: Added GPU autofill for bug reports and implemented per-song loudness-based automatic volume plus saved sing settings.
• 2026-11-13 — Updated Hikari-Tsai/plugin-analyzer: Added an MCP stdio server so LLM clients could inspect, list, and measure plugins safely with new docs and CI smoke tests.
• 2026-11-12 — Updated Hikari-Tsai/ktv-charts: Refreshed song metadata, audio features, vocal ranges, and KTV charts with new track entries and daily rankings.
• 2026-11-12 — Updated Hikari-Tsai/ai-interview: Refreshed interview-answer metadata and updated several AI training/evaluation prompts to the latest source commit.
• 2026-11-12 — Updated Hikari-Tsai/plugin-analyzer: Added a headless CLI and JSON API for plugin measurements, with CI checks, docs, and metrics output support.

_{Last updated: 2026-11-13 16:13 (UTC+8)}

Public repositories    38
Admin private org repos 6
Total stars             135
Last-year activity     742 contributions
Merged pull requests   65 public PRs
Current achievements   Galaxy Brain, Starstruck, Pair Extraordinaire, Public Sponsor, YOLO, Pull Shark x2, Quickdraw
Highlight              GitHub Developer Program Member

_{Last updated: 2026-11-13 16:13 (UTC+8)}

Guardian	500 Days Badge	Dynamic Programming II	Algorithm II
Data Structure II	Programming Skills II	LeetCode 75	Graph Theory I
Show 46 more badges

Area	Technologies
Core Engineering
AI & ML
Web
Audio & Music AI
Plug-in Development
Automation & Platforms

Building practical AI and audio software for creators.