Matthew Liu — I build the systems that keep intelligent software honest.

Working across AI evaluation, observability, and infrastructure, I turn ideas and research into production services built to scale, endure, and evolve.

Path

  1. Johns Hopkins University

    Foundation

    B.S. in Computer Science, grounded in systems thinking, research, and building.

  2. AWS SageMaker

    AI Platforms

    Production model-customization workflows spanning preparation, training, evaluation, and deployment.

  3. Weights & Biases · Weave

    Evaluation and observability

    Systems that make agent behavior inspectable and easier to improve.

Projects

  1. Monopole

    Now · Product direction, evaluation design, and implementation

    A local evaluation workspace for reviewing completed agent runs and turning evidence into better future behavior.

  2. Weave Agent Adapter

    2026 · Architecture, integrations, and observability

    An adapter that turns agent activity into inspectable traces across multiple coding environments.

  3. LLM-as-a-Judge

    2025 · Evaluation UX, judging architecture, and implementation

    Evaluation work that turns long agent traces into structured, evidence-linked assessments using raw context, compact digests, and anchored verdicts.

  4. Model Customization Assistant

    2025–26 · Agent workflow and product engineering

    An agent-guided SageMaker workflow that turns a use case into data preparation, fine-tuning, evaluation, and deployment steps.

  5. UCredit

    2021 · Frontend engineering and product development

    A collaborative advising dashboard built on uCredit to help students and advisors reason about academic plans.