Hi, I'm Morris.

A doctor of pharmacy — with time insideBig Pharma, the hospital, the regulator's table, the lab.

I realized I wanted something bigger —more exciting, with the steepest learning curve I could find.

Nothing today is steeper than AI.Nothing reaches further than a good business decision.

I wanted in. So now I build.

scroll ↓

01 — Introduction · Taipei

Pharm.D. by training,strategist and builderby practice.

A doctor of pharmacy who chose to build — from the world of healthcare and pharma, through the strategy deck, to the codebase.By day I run AI operations at Shopee, the e-commerce arm of Sea (NYSE: SE).By night I DJ to 88,705 people.

Get to know me See the work

25.0330° N, 121.5654° E
EST. 2025 — ONGOING

02The intersection

Most people have one of these.I work where they overlap.

A doctorate in pharmacy taught me to reason from evidence.Big Pharma and a BCG-taught program taught me to find the lever.Now I build with AI. Shopee is where it scales; the projects below are my own.Even the DJ account runs on data: 88,705 followers, 32,353,082 views, 223 days.None of it is rare alone. The intersection is.

"Trained and proven at"

From clinical pharmacy to AI operations.

Hover a logo for what I did there.Tap a logo for what I did there.

Full history on LinkedIn

03What I'm building toward

Everything I've done has been led by curiosity.

It's the one thread through all of it — pharmacy, strategy, the code I write now, the music. I've never been good at just reading about something interesting. I take it apart, find the problem worth solving, and more often than not build something to solve it — which is most of what you'll find on this site.

Right now that pulls me toward the edge where emerging tech stops being a demo and starts changing how people live — robotics, brain-computer interfaces, AI, biotech. Different faces of one question: what becomes possible next, and who gets to build it. I grew up imagining that future; I'd rather be inside it than watching from the outside.

So I look for where that frontier meets something big enough to matter — and where my edge is the overlap: someone who can reason about the system and still ship the thing. I keep building, and I keep following the curiosity until it lands on the problem worth going all-in on.

If something here sparked yours, I'd like to hear from you.

Get in touch

04Selected work

Things I built becauseI wanted them to exist.

Real-time & creative(2)

PulseFrame

Solo build · 2026 · Open source · MIT

Problem: DJs play the music; a VJ runs the visuals behind it. VJ software costs money, and you can only customize it so far. I wanted the freedom to design my own: a visual layer built around my mixes.
Solution: A browser engine that picks beats, energy and BPM out of live audio and drives every scene in real time — no track metadata, no timeline.
Why it matters: Hand-written from raw audio DSP (reading the sound wave itself) to the WebGL scenes (3D graphics in the browser). MIT open source, and already running my own sets.

Case study

Beatbox Viz

Solo build · 2026 · Prototype · actively evolving

Problem: I've been a beatboxer for over ten years — making music with my mouth. I wanted to make that language visible: an interactive toy that turns the sounds I make into something you can see.
Solution: The mic catches each beatbox hit — kick (B), hat (T), snare (K) — and classifies it within milliseconds. A letterform for that sound bursts onto your webcam feed in 3D, timed to the hit.
Why it matters: Where PulseFrame listens to music, this listens to you. Onset detection tuned to 50–80ms — the threshold where the sound and the letter feel like a single event.

Case study

AI systems & automation(2)

Morris Brain

Architect + operator · 2026 · Case study

Problem: Notion-style tools have three problems: the filing and linking never stop being your job, your notes sit on someone else's servers, and an AI can only reach them through a rate-limited API. I wanted the opposite: my context in plain text, on my own machine, maintained for me.
Solution: An Obsidian vault run by AI agents in the terminal (Claude, Codex): I supply documents and questions; they file, cross-reference and maintain everything under written rules — a personal database I can search and cite, down to the source.
Why it matters: The point is compounding: every document filed gives my agents more relevant context to work from, because they run on the same knowledge base I do. Months of unattended operation — and the rules it runs on were adversarially reviewed before I adopted them.

Case study

Finance Research OS

Solo build · 2026 · In the lab

Problem: My positions live across several brokerage apps, and my research and industry work lives somewhere else entirely. I wanted one place that held all of it — the theses, the sector work, and the holdings they're about.
Solution: A nightly pipeline pulls prices and news for every position, then reads it against the thesis I've written for each — flagging what's noise and what actually moves the argument. Everything files into the same knowledge base, where an AI agent can pull it up when I want to dig deeper.
Why it matters: It's run unattended for months — the reading happens overnight, so my attention goes to the decisions, not the upkeep. A de-identified version, with no real holdings, is headed for GitHub — so anyone can run the same setup and manage a portfolio alongside their own AI agents.

Case study

Tools & interfaces(2)

Whisper Scribe

Full-stack, solo · 2026 · Solo build

Problem: Most transcription tools upload your audio to the cloud — and with the alternatives out there, I couldn't be sure what to trust. So I wanted to build my own, end to end, and see how far I could take it solo.
Solution: A local transcription app: drop in a recording, get a timestamped, editable transcript with speaker labels — Whisper inference, fully offline.
Why it matters: Not a novel idea — plenty of these exist. The point was shipping one solo: a working full-stack tool, debugged on real audio across macOS and Windows.

Case study

This Website

Design + build · 2026 · Building in public

Problem: My audience lives on rented platforms — Instagram, YouTube, SoundCloud — and the rest of me is scattered across side projects and half a dozen interests. I wanted one place I own, where every side of me — the DJ, the work, the things I build — adds up to a whole I can share.
Solution: Designed and built as a directed collaboration with an AI coding agent — from reference studies to a hand-tuned design system. Next.js, a custom token system, no off-the-shelf UI kit.
Why it matters: The page you're reading is the proof. Every color, easing curve and shader here was argued over — this is how I build now: taste directing an AI, fast.

Case study

Strategy & leadership(1)

Consulting Casebook

Project manager · 2024 · Published · Free

Problem: Taiwan's aspiring consultants trained on foreign casebooks that never quite fit.
Solution: I led an eleven-person team to write and publish the first local one: 133 pages, nine original cases, vetted by working consultants.
Why it matters: Free to anyone. It's the strategist half of my story, actually shipped.

Case study

Real-time & creative

PulseFrame

Solo build · 2026 · Open source · MIT

Problem: DJs play the music; a VJ runs the visuals behind it. VJ software costs money, and you can only customize it so far. I wanted the freedom to design my own: a visual layer built around my mixes.
Solution: A browser engine that picks beats, energy and BPM out of live audio and drives every scene in real time — no track metadata, no timeline.
Why it matters: Hand-written from raw audio DSP (reading the sound wave itself) to the WebGL scenes (3D graphics in the browser). MIT open source, and already running my own sets.

Case study

Real-time & creative

Beatbox Viz

Solo build · 2026 · Prototype · actively evolving

Problem: I've been a beatboxer for over ten years — making music with my mouth. I wanted to make that language visible: an interactive toy that turns the sounds I make into something you can see.
Solution: The mic catches each beatbox hit — kick (B), hat (T), snare (K) — and classifies it within milliseconds. A letterform for that sound bursts onto your webcam feed in 3D, timed to the hit.
Why it matters: Where PulseFrame listens to music, this listens to you. Onset detection tuned to 50–80ms — the threshold where the sound and the letter feel like a single event.

Case study

AI systems & automation

Morris Brain

Architect + operator · 2026 · Case study

Problem: Notion-style tools have three problems: the filing and linking never stop being your job, your notes sit on someone else's servers, and an AI can only reach them through a rate-limited API. I wanted the opposite: my context in plain text, on my own machine, maintained for me.
Solution: An Obsidian vault run by AI agents in the terminal (Claude, Codex): I supply documents and questions; they file, cross-reference and maintain everything under written rules — a personal database I can search and cite, down to the source.
Why it matters: The point is compounding: every document filed gives my agents more relevant context to work from, because they run on the same knowledge base I do. Months of unattended operation — and the rules it runs on were adversarially reviewed before I adopted them.

Case study

AI systems & automation

Finance Research OS

Solo build · 2026 · In the lab

Problem: My positions live across several brokerage apps, and my research and industry work lives somewhere else entirely. I wanted one place that held all of it — the theses, the sector work, and the holdings they're about.
Solution: A nightly pipeline pulls prices and news for every position, then reads it against the thesis I've written for each — flagging what's noise and what actually moves the argument. Everything files into the same knowledge base, where an AI agent can pull it up when I want to dig deeper.
Why it matters: It's run unattended for months — the reading happens overnight, so my attention goes to the decisions, not the upkeep. A de-identified version, with no real holdings, is headed for GitHub — so anyone can run the same setup and manage a portfolio alongside their own AI agents.

Case study

Tools & interfaces

Whisper Scribe

Full-stack, solo · 2026 · Solo build

Problem: Most transcription tools upload your audio to the cloud — and with the alternatives out there, I couldn't be sure what to trust. So I wanted to build my own, end to end, and see how far I could take it solo.
Solution: A local transcription app: drop in a recording, get a timestamped, editable transcript with speaker labels — Whisper inference, fully offline.
Why it matters: Not a novel idea — plenty of these exist. The point was shipping one solo: a working full-stack tool, debugged on real audio across macOS and Windows.

Case study

Tools & interfaces

This Website

Design + build · 2026 · Building in public

Problem: My audience lives on rented platforms — Instagram, YouTube, SoundCloud — and the rest of me is scattered across side projects and half a dozen interests. I wanted one place I own, where every side of me — the DJ, the work, the things I build — adds up to a whole I can share.
Solution: Designed and built as a directed collaboration with an AI coding agent — from reference studies to a hand-tuned design system. Next.js, a custom token system, no off-the-shelf UI kit.
Why it matters: The page you're reading is the proof. Every color, easing curve and shader here was argued over — this is how I build now: taste directing an AI, fast.

Case study

Strategy & leadership

Consulting Casebook

Project manager · 2024 · Published · Free

Problem: Taiwan's aspiring consultants trained on foreign casebooks that never quite fit.
Solution: I led an eleven-person team to write and publish the first local one: 133 pages, nine original cases, vetted by working consultants.
Why it matters: Free to anyone. It's the strategist half of my story, actually shipped.

Case study

05After hours