AI products · HCI research · Web

I build AI products that help people speak, learn and connect.

I'm Ray Tsai, in the AI master's program at NTU CSIE (HCI Lab). I started in speech-recognition research, shipped an AI speaking-practice app on my own, and build websites for communities and programs.

Ray Tsai giving a thumbs-up in front of the Google sign and colorful lobby wall
NTU CSIE · HCI Lab
MirrorLive on the App Store

From speech research to shipped products

Research taught me how to measure; products taught me what to cut. Here is where I study and what I have done.

Education

National Taiwan University

Dept. of Computer Science & Information Engineering — AI master's program, HCI Lab

Research

2025/6 – 2026/3

CKIP Lab, Institute of Information Science, Academia Sinica

Part-time research assistant

Speech recognition challenge

2nd

Track 1 — 2nd place

3rd

Track 2 — 3rd place

Formosa Speech Recognition Challenge 2025

Team entry (student division); the team's paper was published at ROCLING 2025.

Summer 2026

≈7weeks

NTU Silicon Valley exploration program

About seven weeks in Silicon Valley doing market research and expert interviews, followed by the Demo Day on Sep 12.

Read the seven-week recap (opens in a new tab)
Group photo at night
Group photo at night
Group photo at the office
Group photo at the office
Sunset at Yosemite
Sunset at Yosemite
Sharing at an office in Silicon Valley
Sharing at an office in Silicon Valley
Group selfie at Google
Group selfie at Google

Film · YouTube · 10:36

I turned my seven weeks into a film

Seven Weeks in Silicon Valley | Growing Into My Own Voice

YouTube loads only after you press play.

Watch on YouTube (opens in a new tab)

Turning research and curiosity

into tools people use daily

A shipped app, and products in progress

Mirror home screen with pixel character
Mirror voice fingerprint result

Live on the App Store

Mirror – AI Speech Coach

An AI speech coach: record your spoken answer, and Mirror transcribes, scores and rewrites it for clarity and structure — then plays the better version back in your own cloned voice.

  • Voice signature in 30 seconds
  • Mock interviews in English, Chinese, Japanese
  • Side-by-side replay of original vs. improved
  • Five-axis voice radar

Built solo · iPhone · Free

View on the App Store(opens in a new tab)

More projects

Open a project to see its screens, status and links.

Echo iOS · Chrome
Echo product page and a lock-screen practice prompt

A podcast player that remembers every 15-second rewind: it keeps the sentence you replayed, diagnoses why you missed it (vocabulary, linking, speed…), and turns it into tomorrow's speaking practice. The Chrome version does the same with YouTube captions.

  • iOS: invite-only TestFlight beta
  • Chrome extension: in development
Rayality Projection Mapping Web · Open source
Rayality editor corner-pinning and projector output

Projection mapping in the browser: import images or video, drag four corners to correct perspective, and send the output to a projector from a separate window. No API key needed for the core workflow; optional Google Veo generation with your own key.

  • Live · open source
MedBuddy Web · LINE bot
MedBuddy caregiver dashboard (demo data)

A prototype for medication understanding and care handoffs for older adults on many medicines: rule-based checks, and two LINE accounts sharing one care record between elder and caregiver. Medication-bag photo readings must be confirmed by a person before they are saved. Built in about 48 hours during a build challenge.

  • Prototype · demo data
Echo product page and a lock-screen practice promptRayality editor corner-pinning and projector outputMedBuddy caregiver dashboard (demo data)

Websites I've built

Web design service

A website for your shop or studio

Student pricing for simple showcase websites for small shops, studios and individuals. Limited slots each month.