EXPERIENCE / RESUME

Linsama

Software Engineer · Developer Platforms, DevOps & AI-assisted Development

I have an engineering background in developer infrastructure, delivery platforms and distributed systems. My recent work includes LLM-assisted workflows, alongside independent web products, games and agent projects.

Seeking software engineering opportunities in Japan or Canada, with a focus on developer platforms, backend systems, and AI applications. I’d be happy to connect about roles that offer work visa sponsorship.

Profile & contact · LinkedIn ↗GitHub ↗

A NOTE ABOUT ME

From developer tools to products of my own.

Much of my career has been in developer infrastructure: delivery pipelines, feature flags and artifact repositories, alongside architecture and team development. That work shaped how I approach software: how a system runs, how problems can be diagnosed, and how a team keeps delivering.

I now bring that engineering background to AI-assisted development and personal products. I explore agents that act on sustained goals, and I build games people can try. I care about the work after the first version, too: defining boundaries, checking results and iterating through actual use.

If your team is building developer platforms, exploring AI workflows or taking an early product toward delivery, I would welcome a conversation about the engineering and collaboration involved. The experience and projects below offer a view of how I work.

Work experience

Nov 2025 — Present

China Merchants Bank NT

Senior Software Development Engineer

Audit orchestration: Developed an asynchronous engine with dynamic upstream integration, combining static operators and LLM-driven evaluation to organize complex audit rules and scenarios.
AI-assisted coding: Built a specification-driven workflow for self-hosted models with limited context, using explicit file and task state to organize development into stages.

Jan 2025 — Nov 2025

Independent Developer

Founder / Full Stack Developer

Built an AI news SaaS product with a Next.js frontend and Python ETL pipeline for LLM-based classification, summarization and daily delivery.

Mar 2020 — Dec 2024

FAW-Volkswagen

Architect / Team Leader

Continuous delivery: Established practices across backend, frontend, mobile, AI and vehicle software, incorporating code management, security checks and automated testing.
Developer platform: Built an infrastructure-as-code platform integrating Jira, Gerrit and Kubernetes, connecting requirements, builds and deployment with visibility into quality and delivery metrics.
Team development: Mentored engineers through code reviews, technical sharing, agile practices and regular one-on-one meetings.

Oct 2017 — Jan 2020

Meituan

Software Development Engineer II · Infrastructure

Feature-flag SDK: Developed a Java SDK that synchronizes configuration asynchronously and evaluates flags in memory, separating runtime decisions from remote configuration retrieval.
Artifact repository: Led the distributed refactoring of an internal Maven repository, improving metadata serialization and multi-level caching to address artifact-access performance.

Feb 2017 — Oct 2017

Microsoft

Software Development Engineer I

Developed C# business logic for Bing Ads using internal data-processing infrastructure.

Jul 2015 — Feb 2017

Baidu

Software Development Engineer I

Designed a pipeline-stage-job execution model and improved distributed task scheduling for a developer delivery platform.
Adapted delivery tooling for hybrid and private cloud deployments.

SELECTED PROJECTS

Ideas you can try.

These projects extend my interest in system design while bringing product, interaction and release decisions into the picture. I use AI assistance in development and asset creation, while driving product direction, technical choices, review and acceptance.

01 / Python / SQLite / MCP

Nagi · An agent with ongoing projects

An agent project in development, exploring how an agent can act on sustained interests and projects, with a Live page showing its current activity.

Visit Nagi Live (in development) ↗

My work: Organizing schedules, activities, projects, memory and tool use into a running system, with web pages for observation and communication.
Technical choices: Separate activity selection from capabilities: the loop schedules work, while capability modules supply tools and context. Code maintains project state, with explicit evidence requirements for completion and review.
How it is checked: Activity logs, task state and separate review sessions make progress inspectable. Interrupted-work recovery and failure backoff are part of the runtime. The project remains an ongoing exploration.

02 / Cocos Creator / TypeScript

Eight Wastes Survivors · Reproducible combat

A survival game inspired by Chinese mythology, bringing movement, automatic attacks, upgrade choices and boss encounters into a complete run.

Play Eight Wastes Survivors ↗

My work: Developing gameplay rules, balance configuration, combat feedback and channel builds, with characters, weapons, items and enemies represented as configurable content.
Technical choices: Use Cocos Creator and TypeScript, separating combat rules from rendering. Fixed simulation steps, seeded randomness and recorded input support replay, so rule behavior can be inspected independently of the visuals.
How it is checked: The codebase includes rule tests, replay tests and batch simulation tools. These support logic checks, while playtesting informs pacing and interaction feedback.

03 / Three.js / TypeScript / Cloudflare

Crazy Goose · A game on tiny 3D planets

A lighthearted game about chasing, growing and causing trouble, connecting 3D environments, progression and daily challenges.

Play Crazy Goose ↗

My work: Developing movement on a sphere, interface interactions, progression and save flows, alongside web and mini-game platform adaptation.
Technical choices: Render the world with Three.js and separate pure TypeScript rules from rendering and platform APIs. The client and server share daily-challenge rules; the leaderboard uses Cloudflare Workers and D1.
How it is checked: Tests cover rules, save migration, language selection and Canvas text wrapping. A bot helps investigate balance, while playtesting checks the interaction experience.

Let’s talk.

For my full professional profile, more about my experience, or work and collaboration opportunities, visit my LinkedIn.

Profile & contact · LinkedIn ↗