One-person Staff-Level Advisory, Est. 2019

Calm, rigorous machine learning and data work for consequential teams.

I help teams turn ambiguous data and ML problems into robust systems, sharper decisions, and measurable outcomes — without theatrics, office politics, or performative process.

Review My Engagement Terms →

Currently

Booking Q4 2026

Based

Taiwan — Working Globally

Engagement Type

Fully Remote; Async-first

The practice

Veritable

A single senior practitioner, chosen deliberately.

Most ML consultancies are agencies aiming for scale, billable hours, and overhead. I run a single-person office for an entirely different reason: intellectual challenge and tangible impact.

Having built a comfortable financial cushion from my previous engagements, I consider only new engagements in which I can put my expertise and working style to full use. Every client gets my full attention, my honest opinion, and a direct line of communication that never goes through a project manager.

What I help with

High-signal ML leadership without the hype.

01 →

Production ML Engineering

I step into half-built systems and turn them into robust production systems. Drawing on staff-level experience building core ML infrastructure at a category-leading vertical online retailer with $400M+ in annual revenue, I design and ship systems that perform under real commercial constraints.

  • Core ML infrastructure design
  • System rescue & modernization
  • Pragmatic build-vs-buy decisions
02 →

Modeling That Holds Up

Once ranked 63rd globally on Kaggle, I've spent years solving hard modeling problems in competitive, leaderboard-ranked environments. From CV, tabular, and recommender systems to pragmatic LLM and RL post-training work, I build models optimized for actual business metrics, not vanity benchmarks.

  • CV, tabular & recommender systems
  • Evaluation & metric design
  • LLM & RL post-training
03 →

Fractional ML Leadership

I provide staff-level clarity for your AI strategy, grounded in a deep computer science and statistics background. I help technical founders set direction, establish rigorous evaluation methodologies, and debug hard-to-explain model behavior.

  • ML roadmap & technical direction
  • Technical due diligence
  • Statistical debugging & validation

How I work

A small set of beliefs, held consistently.

  • 01

    Remote and async by design.

    I work fully remotely from Taiwan and operate across time zones without friction. No on-site visits, no performative face time. You get focused output delivered in writing — not presence billed by the hour.

  • 02

    Written, not performed.

    The best deliverable is a clear technical memo. Fewer slides, more thinking made visible in code and documentation. If a decision matters, it should survive being read, not just heard in a meeting.

  • 03

    Evidence first, consensus second.

    I'll tell you what the data says and where your system is breaking — plainly, with evidence. Then I'll help you decide what to actually build. I don't optimize for consensus; I optimize for clarity.

  • 04

    One senior, end-to-end.

    You hire me. Not a pyramid. The person doing the work is the person writing the code, reviewing the math, and owning the trade-offs. No handoff to a junior associate you've never met.

  • 05

    Pragmatic over hyped.

    I am allergic to model-bloat. If a well-specified statistical model or a simple ensemble achieves the outcome at a fraction of the cost and complexity, that is what we build. Sophistication is a tool, not a goal.