Martin Lu

Seattle, WA · Berkeley, CA

Martin Lu

Mathematician and software engineer. I build reliable systems and expressive tools.

GitHubLinkedInResumeEmail

about

I studied math at UC Berkeley and write software at Amazon Fulfillment Technologies. On the systems side I build distributed backends and full-stack services: async read-through caching, real-time telemetry at 20k requests per second, React and Redux on the frontend. On the research side I developed a custom one-shot symbol detection computer vision model for Quotr.ai: custom model architectures, the synthetic datasets to train them, loss functions, and optimization algorithms.

work

  1. Amazon

    Bellevue, WA

    Aug 2025 — Present

    Software Development Engineer, Fulfillment Technologies

    Built a full-stack configuration service with DynamoDB read-through caching, Redis, and a React/Redux frontend, used across 26 fulfillment centers. Cut operator configuration time by 78%.

  2. Quotr

    Berkeley, CA

    May 2025 — Aug 2025

    AI Research Software Engineer Intern

    Built a one-shot symbol-recognition engine (cross-attention with template matching, Conformer backbone) at 80% mAP and 1.5s P99, served from a FastAPI endpoint on EC2. Generated 500k synthetic document-symbol pairs that raised recall 20 points and removed 1,500 hours of manual labeling.

  3. Amazon

    Seattle, WA

    Jun 2024 — Aug 2024

    Software Development Engineer Intern

    Designed a fulfillment-center telemetry service handling 20k requests per second on EC2, ALB, and DynamoDB. Wrote the Spring/Smithy REST API and provisioned everything with CloudFormation, cutting deploys from 1 hour to 10 minutes.

  4. Headway Technologies

    Milpitas, CA

    Jun 2023 — Oct 2023

    Software Engineer Intern

    Formulated a MILP batch-scheduling model from operations-research literature; simulated on 12 months of fab data and predicted an 8% queue-time reduction. Built eight Spotfire dashboards for line managers.

  5. Phenomena Learning

    Seattle, WA

    Jun 2022 — Sep 2022

    Full-Stack Developer Intern

    Shipped four UX features and a 2D animation engine in Vue.js that let non-technical educators build animations without code. Cut lesson-creation time by 25%.

  • UC BerkeleyB.A. Mathematics · 3.81 GPA, 4.0 in major · 2023 — 2025
  • UC San DiegoB.S. Applied Mathematics, Minor in Computer Science · 2021 — 2023

projects

Quotr Symbol Engine

ML

One-shot computer vision for unseen architectural symbols. A cross-attention model recognizes new symbol classes from a single example at sub-second latency.

proprietary

WebStrigil

ML

An LLM-guided recursive web crawler that decides where to go next, built on Playwright and Scrapy.

GitHub

Fulfillment Telemetry Service

Systems

A real-time analytics service processing 20k package movements per second on AWS, feeding live efficiency metrics to fulfillment-center operators.

proprietary

Weplate

Systems

A React Native nutrition-tracking app for students. I led the frontend, with Redux and Recoil state and a typed mock API for parallel development.

GitHub

Spanimate

Creative

A multi-scale animation editor for p5.js sketches. Edit the same drawing as code or by direct manipulation, at several levels of abstraction.

GitHubWrite-up

Evolution Simulator

Creative

Physics-driven creatures controlled by small neural networks that evolve over generations.

GitHub

writing