Hank Lin

Engineering Leader · Panasonic Avionics

Hank Lin

Senior Manager, Machine Learning · Panasonic Avionics

I lead teams building large-scale ML and personalization systems that serve millions — and I still ship code across mobile, web, and systems. A Ph.D. researcher at heart, an engineering leader by trade.

LinkedInGitHubRésumé

About

Engineering leader, 10+ years

Engineering leader with 10+ years building large-scale personalization, ranking, and ML systems that serve millions. I take teams from a blank page to launch — setting technical direction, coaching engineers and scientists, and staying close enough to the code to own the hard calls.

My work is grounded in science — a Ph.D. in Computer Science and years deep in machine learning and NLP, from location-aware ranking to Alexa book Q&A to modeling character relationships in fiction. And I'm still building at the edge: a Go engine written from scratch — a hybrid conv-transformer trained on 12.6M positions, wrapped in a cross-platform Rust engine that plugs into the standard Go tools. (Formerly Hanlung Lin.)

I build across platforms, too. I ship software end to end — mobile games in Godot, cross-platform apps in Flutter and SwiftUI, full-stack web on AWS serverless, and systems code in Rust and Python — and I like owning the whole path from idea to production.

Side projects

Updated 2026-08-15

3,703 contributions in the last year, built in my spare time

  • Ph.D., Computer Science — University of California, Riverside
  • M.S., Computer Science — University of Electro-Communications, Japan
  • B.S., Computer Science — University of Electro-Communications, Japan

Side Projects

Projects

Built in my spare time — the repos behind the graph above.

Seven Seas Kingdom

A pirate idle-strategy mobile game — command fleets, raid and conquer islands, and grow a home port that keeps earning while you're offline.

  • Multi-ship fleets and a multi-area empire with away-area simulation
  • Automated wave combat with tactics, loot, and crew/gear/forge systems
  • Deterministic seeded simulation, headless-testable (~1,957 tests)
  • Reducer-store architecture over a reusable game-agnostic SimEngine
Godot 4.7GDScriptiOS / Android

Go Player Profiler

Reverse-engineers a Go player's style from their game records — runs KataGo on every game, classifies each move, and surfaces the shapes and board zones they lean on.

  • Five-stage pipeline: fetch → ingest → derive → report → scout
  • Desktop GUI with unified search over online and local game libraries
  • PDF style reports; bilingual JA/EN interface
PythonKataGoPyInstaller

Large Go Model

A Go engine built from scratch — a hybrid conv-transformer ("HybridNet") trained on 12.6M positions, served by a cross-platform Rust playing/analysis engine that runs in standard Go tools like Sabaki and KaTrain.

  • Data pipeline over 51k games → 12.6M distillation positions
  • PyTorch training stack (muP/Muon, ONNX export)
  • Rust engine speaking GTP and KataGo analysis (works in Sabaki / KaTrain)
PyTorchRustONNXGTP

InsectWeb

A kid-safe family insect blog with a real social feed — photo diaries, threaded comments and reactions, auto-moderation, and a species guide. Live in production at insect.ninja.

  • Threaded comments with per-comment reactions and auto-toxicity moderation
  • Cognito auth with admin / parent / kid roles
  • Species guide with photo identification; auto image resizing to WebP
  • Live at insect.ninja
Next.js 15TypeScriptAWS CDKLambda / DynamoDB

KnowledgeIndexer

A CLI and MCP server that reads any Git repo into a Source → Feature → Function map with an LLM, then exports one KNOWLEDGE.md that AI agents can navigate.

  • Provider-agnostic summarization (Claude or any OpenAI-compatible endpoint)
  • Diff detection to skip unchanged sources
  • Ships an MCP server exposing index/list/show/query/export as tools
TypeScriptNodeMCPAnthropic SDK

Engineering Method Skills

Seven battle-tested engineering-method skills for AI coding agents — evidence discipline, systematic debugging, design-before-build, and change control, each shipping with its own trigger evals.

  • Each skill ships a SKILL.md plus a trigger eval set
  • Drawn from a real 2,000-commit game repository
  • Encodes how to verify, debug, design, and keep docs honest
Skill / MCP authoringAI agents

Expertise

Skills

Leadership

Engineering LeadershipTeam Building & MentorshipProduct Strategy & RoadmappingCross-Org Influence

Science

Machine Learning & Deep LearningDeep Learning ResearchRecommender & Ranking SystemsNatural Language Processing

Platforms & Languages

Mobile — Godot, Flutter, SwiftUIWeb — Next.js, AWS ServerlessSystems — RustPython · TypeScript

Career

Experience

  1. Senior Manager, Machine Learning

    Panasonic Avionics Corporation

    Dec 2025 – Present · Irvine, CA

    • Lead four engineering teams across the in-flight connectivity platform — ground operations, cloud, air-to-ground communications, and air-side log aggregation.
    • Own the machine-learning charter for the connectivity stack, from aircraft-side data to cloud services.
    • Set the technical direction and grow the teams spanning aircraft-to-ground and cloud systems.

  2. Software Development Manager

    Amazon — Personalization

    Feb 2024 – Dec 2025 · Irvine, CA

    • Set the vision and strategy for interest- and product-concept personalization powering Amazon retail.
    • Led 8 engineers and 2 applied scientists shipping ML services to millions of customers worldwide.
    • Partnered across product, science, and platform orgs to keep it reliable, scalable, and reusable across domains.

  3. Sr Software Development Engineer

    Amazon — Personalization

    Mar 2021 – Feb 2024 · Irvine, CA

    • Drove the Cost-to-Serve personalization initiative across multiple organizations.
    • Designed location-aware ranking components that sharpened regional relevance.
    • Managed and grew a team of 4; coached multiple promotions.

  4. Sr Software Development Engineer

    Amazon — Kindle

    Apr 2018 – Mar 2021 · Seattle, WA

    • Led book-specific Question & Answer for Alexa, end to end.
    • Built intent-classification models to read natural-language book questions.
    • Built an NLP system that maps character relationships across fiction.

  5. Software Development Engineer

    Amazon — Prime

    Jan 2015 – Apr 2018 · Seattle, WA

    • Engineering point of contact for the Amazon Fresh integration.
    • Built a simplified global Prime subscription used by millions of customers.
    • Supported Prime launches across India, China, and the EU.