DATA SCIENTIST · AI ENGINEER · EX-UNDERWRITER · SINGAPORE
I'm Ethan Mak. These days I build agentic AI systems; earlier in my career I trained the models myself. Either way it ends the same: wrapped in an API, running on infrastructure I manage — from a property-valuation engine that served 150,000 requests a month to an agentic platform where one request sets a team of AI agents to work.
/ THE SHORT VERSION
I like owning the whole stack — model, API, server. This site is the proof: the work demos, the AI media tools, the game, and the blog all run on a machine I manage myself.
2019 — NOW
01
A hub orchestrator for federated AI agents: one chat interface, natural-language requests routed by a lead agent to specialised member agents, results streamed back in real time. Existing services plug in over REST, MCP, or A2A, and a workflow builder turns routine requests into reusable tools. I lead the project — architecture, build, and delivery.
02
A meta-harness: the private control plane I built to run a company of coding agents on real projects. Eighteen roles plan, build, independently review and deploy, while the app owns approvals, sandboxes, crash recovery and Git integration. A chat reply never counts as done. Built solo.
03
An insight engine for large document collections. Point it at thousands of texts and it surfaces the themes, patterns, and signals that matter — turning weeks of reading into minutes of analysis. In production with 4,000+ users across multiple organizations and 60+ million documents analysed.
04
A property-valuation API for Singapore's non-landed private and MOP HDB market, serving 150,000+ requests a month. Cut valuation error from ~5% to ~2% while lifting coverage from 88% to 98% — live in production at a Singapore property portal.
05
The machine behind everything here: a self-managed server running a music streamer, browser-based AI image and video tools, a rhythm game, and this very site — blog included. If it's on ethan1989.com, it lives here.
/ FROM THE BLOG
Occasional posts on projects and what building them taught me — short, personal, and written to be read, not skimmed.
26 SEP 2026
How I built Illustrious, a private control plane that runs a company of coding agents, and why the layer above the harness decides what counts as done.
30 JUN 2026
Why my team stopped building single-purpose AI tools and built an orchestrator instead — the story behind the agentic platform on my work page.
25 DEC 2025
A semester embedded with a SaaS company's sales pipeline — multi-touch attribution, production-grade lead scoring, and finding where sales effort actually pays.
/ OFF THE CLOCK
RHYTHM GAME
A Patapon-inspired rhythm-command game set in the deep sea — built from scratch, every sound synthesized in the browser. Phase one of a long-term hobby.
Play it →
SELF-HOSTED AI TOOLS
Generate and edit images, create H3 or Seedance videos, and assemble longer clips on a timeline. Bring your own fal.ai API key and pay only for what you use.
Try them →
LEARNING PROJECT
A visual, interactive revision platform for sharpening the algebra, calculus, and statistics behind data science.
Explore it ↗
AWS CERTIFICATION
A source-backed learning platform for AWS Solutions Architect certifications, with interactive lessons, original assessments, and progress tracking.
Start learning ↗
/ BACKGROUND
Seven years shipping production ML and AI systems — and five years in insurance before that, across underwriting, reinsurance, and risk. I've been on both sides of the model.
2013 — 2016
D&O and professional liability up to USD 10M limits; 200+ risk assessments across industries.
2016 — 2017
Facultative placements for cedents across 6 APAC countries.
2017 — 2018
ORSA, risk appetite, and board-level risk reporting.
2019 — 2022
Production property-valuation models powering live pricing at scale.
2022 — NOW
Agentic AI and text-analytics platforms; leading a 4-engineer team.
EDUCATION
Computational Data Analytics track — deep learning, NLP.
/ TOOLBOX