Game concepts with systems depth
Project Sanctuary/cardlyB and Jianghu Ascendant show two production paths: a Godot 4 .NET roguelike card battler and a browser-based wuxia deckbuilder with a live Three.js battlefield.
Playable games / AI platforms / agent infrastructure
Vincent Lin / adroitous
A GitHub-first portfolio of fundable game concepts, full-stack AI products, and developer harnesses for agent-heavy production. The through line is practical systems work: playable surfaces, real pipelines, and tools that make iteration measurable.
- 2 game worlds
- Godot card battler + browser wuxia roguelike
- AI products
- SAT prep, tutors, telemetry, and workflow tools
- 11 departments
- autonomous game studio harness architecture
Why this portfolio exists
Proof of range across games, AI, and production systems.
Agent-native developer workflows
LocalTelemetry, Game Studio Harness, and Claude PowerShell Hooks focus on making coding agents observable, coordinated, and useful during long-running local development.
Full-stack AI education products
AcePrepLabs expands the portfolio into applied AI learning: tutoring, practice tests, flashcards, analytics, mobile training, and question-bank tooling.
Curated highlights
Projects worth opening first.
Live browser game
Jianghu Ascendant
Browser-based wuxia deckbuilding roguelike, playable now at jianghu-ascendant.pages.dev. The build combines a Three.js 3D battlefield with Qi-based martial combat, four martial schools, relics, potions, cultivation progression, and bilingual Chinese / English presentation.
Flagship game project
Project Sanctuary / cardlyB
A narrative-first, anime-styled, party roguelike card battler built in Godot 4 .NET with C#. The local workspace includes a playable implementation, full design bible, generated art assets, BGM/SFX pipelines, mobile planning docs, and QA harness work.
Full-stack AI education platform
AcePrepLabs
Applied AI SAT prep and training ecosystem spanning a web platform, mobile corporate training app, and College Board question-bank tooling. The work demonstrates product breadth: tutoring flows, assessment engines, analytics, mobile delivery, and automation.
Browser demo
JavaScript
Physics Equation Playground
Interactive browser playground for physics and equation-driven canvas animations, built as a fast visual prototyping surface.
Agent tooling
PowerShell
Claude PowerShell Hooks
Windows 11 notification hooks for Claude Code, focused on making long-running local coding sessions easier to monitor.
AI app
HTML/Python
Resume JSON Generator
Job-description-driven JSON Resume generator with Python models, validation, prompts, and a lightweight frontend.
NLP project
Python
YouTube Comments Classifier
NLP project exploring comment ranking and generation using sentiment, co-occurrence features, Markov chains, and recurrent models.
ML experiments
Python
K-Means Experiments
Clustering experiments across many benchmark datasets, with result outputs and analysis scripts for comparing behavior across domains.
Systems/security
C
Buffer Overflow Exercises
Low-level C exercises around memory behavior and exploit mechanics, useful as evidence of systems-level fundamentals.
Harness ecosystem
Infrastructure for agent-heavy work.
AI agent telemetry harness
LocalTelemetry
Local workbench that ingests Claude Code and Codex sessions into SQLite, then exposes the data through Python/FastAPI, a CLI, and a REST API. The goal is source-of-truth telemetry for agent behavior, session replay, and development observability.
Pre-production AI game studio
Game Studio Harness
AI game development framework with 11 autonomous agent departments, Dolt-backed shared memory, and Bash/Python/Claude Code orchestration. It turns game pre-production into coordinated departments rather than one-off prompts.
Build pattern
The work is connected by systems thinking.
Game production loop
Specs define the design contract, data files drive runtime behavior, engines present the experience, and QA scripts close the loop across both Godot and browser-native builds.
- 01 Vision, pillars, and market positioning
- 02 Core simulation, combat rules, and game data
- 03 Art, SFX, BGM, 3D scenes, and UI pipelines
- 04 Playtest, telemetry, QA harness, and iteration notes
Why it matters to funders
This portfolio shows more than isolated demos. It shows a builder who can move from research to product to game implementation, while keeping scope, production constraints, and technical proof visible.
Playable proof
live web game plus Godot card battler work
Product range
edtech, mobile, dashboards, and automation
Technical range
C#, TypeScript, Python, PowerShell, C
Current focus
agent orchestration, observability, and game systems
GitHub first
Open the repos, then follow the strongest threads.
The public GitHub account is the main call to action. It surfaces playable game builds, agent tooling, AI workflow projects, edtech prototypes, and the longer technical history behind the portfolio.