Project archive
Explore two end-to-end products, five applied AI workflows and two open-source utilities, with my role and the evidence behind each.
01 · End-to-end product
FightGame
A personalized-avatar multiplayer pixel RPG with a shared world and synchronized online skill-card battles.
My role: product direction, system boundaries, Agent coordination, review and end-to-end acceptance.
02 · End-to-end product
Niulai Squad
A four-Agent BNB Smart Chain product that turns sourced market signals into a traceable story while keeping voting, execution and publishing independently controlled.
My role: product architecture, operating rules, permission boundaries, failure review, release decisions and iterative full-flow UAT.
Applied AI Workflows
Compact evidence of workflow design, safety boundaries, data provenance and acceptance—without inflating each workflow into an end-to-end product claim.
A guarded two-machine workflow that moves AI video projects from storyboarding and approval through queued generation, monitoring, synchronisation and final rendering.
Workflow direction, approval and cost boundaries, integration decisions, review and delivery acceptance.
Real generation-to-render workflow validated
A simulated reference runtime that turns moderated input into shot plans, generation and quality-control states, playout buffering, replay fallback and auditable budget control.
Product direction, state and safety boundaries, acceptance criteria, verification and limitation review.
Simulation tested · not a live channel
A source-linked document system that turns complex records into searchable timelines, relationships and reviewable evidence while preserving page-level provenance.
Product model, evidence and privacy rules, review workflow, release boundaries and acceptance.
Public framework and live demonstration available
View public framework ↗A static-first data product that transforms multiple official sources into explainable project timelines without confusing observed, estimated and projected values.
Product direction, source and language constraints, cross-Agent review, defect interception and acceptance.
Tested local data-product milestone
A deterministic workflow that turns approved visual layers into quota-controlled NFT collections, marketplace metadata, visual QA reports and safely gated platform staging.
Product direction, collection rules, visual approval, launch parameters, Agent coordination, QA and final acceptance.
888-item package validated · platform draft staged
Open-Source Utilities
Focused tools, released with documentation and explicit limitations.
GitHub data tool · Public release
A GitHub data tool that discovers and ranks fast-growing AI Agent skills and plugins using recent star activity and transparent relevance scoring.
Public release · CI tested
View repository ↗Maintainer CLI · Public release
A CLI that audits documentation, contribution workflow, CI, testing, release hygiene and basic secret safety before a repository is shared publicly.
Public release · Multi-version CI
View repository ↗