Skip to content
Tobias Taulien

Taipei, Taiwan

AI product manager and founder, building from the heart of Asia.

Ten years shipping other people's products. Now I'm building my own in Taipei — along with the demos, notes and tools on this site.

See what I'm building Get in touch
Tobias Taulien

  • 10+ yearsin product

  • Founderof NearBuy

  • NTU GMBAPresidential Award

  • 8 AI demosbuilt and live

currently building

NearBuy

在地好店，一指即達 — local gems, just a tap away

A marketplace connecting shoppers with the independent retailers around them — live around NTU and expanding across Taipei.

since January 2026
Live
where it launched
Around NTU
and growing
Taipei
Read the case studyVisit NearBuy
The NearBuy app, showing local stores and products near National Taiwan University

AI demos & experiments

Play with them. Learn how they work.

All demos

Tokenizer Visualizer

See how an LLM splits your text into tokens.

Tiny Neural Net Playground

Watch a tiny neural network learn, right in your browser.

WebLLM Persona Chat

A real LLM running on your GPU. No API calls.

GPU Rush

An idle game about tokens, GPUs and context windows.

recent notes

Thinking in public.

All notes

Building

Seven Weeks, Seven Tabs: What Happened to My Week Dashboard

· 6 min read

AI

I Finally Understand What 'Attention' Actually Does — and It Changed How I Scope AI Features

· 4 min read

AI

What 4.5 Hours of Karpathy Taught Me About Being a Better Product Manager

· 4 min read

the journey

How I got here.

The longer version

  1. 10+ years

    Product

    BMW, Stock3, Be Around and GoGroup — ending as a Lead Product Manager running client delivery end to end.

  2. NTU Global MBA

    Asia

    Moved to Taipei for the Global MBA at National Taiwan University. Presidential Award, Best of Class 2025.

  3. Since January 2026

    Building

    NearBuy, AI demos built with Claude Code, and notes written in public.

resources

Useful stuff for builders.

All resources

Evals Cheat Sheet

The order of operations I use for evaluating an AI product — from cheap assertions to error analysis, in rough order of cost. Explained in my note on why evals are a PM's job.

↓ Download

PM Workflow Cheat Sheet

The one-pager I actually use at NearBuy — how I run PRDs, data analysis, product strategy, and vibe-coding sessions with AI, step by step. Real workflow, not theory.

↓ Download
doc

Build Your Own Job Board — PRD

A spec you drop into Claude Code to build a personal job search board: a filtered feed from four job APIs, a pipeline for tracking applications, and a CV tailored to one posting. Every number in it was measured against the live APIs, not read from their docs — including the one parameter that took relevance from 4% to 100%, and the API that truncates every description at 500 characters.

↓ Download
9 files

Startup Review Agents

Eight Claude Code subagents I use as a virtual review board — an engineer, an investor, a skeptical customer, and more, each pressure-testing specs, pitches, and content before the real world does. Drop the .md files into .claude/agents/ and Claude Code picks them up automatically.

9 files · Markdown →

Working on something?

Hiring for an AI product role, building something in Taiwan, or just want to talk shop — email me.

Get in touch

or email hello@tobiastaulien.com