I build AI-agent workflows, serverless SaaS, and privacy-first web, mobile, and desktop products.
React · React Native · Expo · Next.js · TypeScript · Node.js · AWS Amplify Gen 2 · Electron · MCP
Open to remote full-stack, React Native/Expo, product engineering, and AI-agent platform opportunities.
- Flowtive (private codebase; public releases and extensions) - Offline-first project management for Kanban, sprints, analytics, LAN sync, and AI-agent collaboration. Available on the VS Code Marketplace, Open VSX, and the Chrome Web Store; public release assets include signed desktop installers.
- Cycles (private) - Human-gated engineering loop. Flowtive tasks flow through Flowtive MCP and OpenClaw to an Engineer that runs every 10 minutes and implements through OpenCode ACP. A Reviewer runs every 15 minutes to gate PRs and route rework back to Flowtive.
flowchart TB
B[Backlog column\nnot auto-claimed]:::column
T[To Do column\nactive sprint]:::column
IP[In Progress column]:::column
RV[Review column]:::column
RF[Review Failed column]:::column
NR[Needs Rebase column]:::column
TF[Test Failed column]:::column
BL[Blocked column]:::column
D[Done column]:::column
B -. Human adds to sprint .-> T
T --> M[Flowtive MCP]
IP --> M
RF --> M
NR --> M
TF --> M
M --> O[OpenClaw and Cycles]
O --> ED[Engineer discovery\nresume owned work or select eligible task\nevery 10 minutes]:::engineer
ED -->|Eligible or rework| IP
IP --> E[Engineer implementation]:::engineer
E --> A[OpenCode via ACP]
A --> P[Pull request]
P --> RV
E -->|Blocker| BL
BL -. Human resolves .-> IP
ED -->|No eligible task and improvement enabled| ID[Engineer improvement discovery\nrate-limited per project]:::engineer
ID --> S[Suggestion column\nhuman promotes to To Do]:::column
S -. Human promotion .-> T
ED -->|No eligible task; no discovery due| EQ[Quiet exit]
RV --> RD[Reviewer discovery\nPR work first\nevery 15 minutes]:::reviewer
O --> RD
RD -->|Reviewable PR| R[Review PR and task evidence]:::reviewer
R -->|Pass| H[Human merge approval]
H --> D
R -->|P0/P1 findings| RF
R -->|Merge conflict| NR
R -->|P2/P3 follow-up| F
RD -->|No PR and scan due| I[Read-only risk scan\nmax once per 4 hours per project]:::reviewer
I --> F[Reviewer finding or P2/P3 follow-up\nfiled in To Do]:::reviewer
F --> T
RD -->|No PR; scan not due| RQ[Quiet exit]
P -. Planned testing path .-> TD[Tester discovery\nPR awaiting verification\ncoming soon]:::tester
TD -. Planned .-> TV[Tester verification\ntest, lint, typecheck, build\ncoming soon]:::tester
TV -. Pass .-> RV
TV -. Fail .-> TF
classDef column fill:#f8fafc,stroke:#64748b,color:#334155;
classDef engineer fill:#dbeafe,stroke:#2563eb,color:#1e3a8a;
classDef reviewer fill:#f3e8ff,stroke:#9333ea,color:#581c87;
classDef tester fill:#f8fafc,stroke:#94a3b8,stroke-dasharray: 5 5,color:#64748b;
Blue nodes are Engineer paths; purple nodes are Reviewer paths; muted, dashed nodes are the planned Tester path. Backlog tasks are intentionally excluded until a human adds them to an active sprint. The Tester role is not active yet; it will verify PR-ready branches and record evidence before work proceeds to review or rework.
- openclaw-migrate - Cross-platform CLI that safely exports, inspects, restores, and rolls back OpenClaw agent environments.
- Atlas Photo - Privacy-first Electron photo application with local EXIF/GPS processing, interactive maps, thumbnail caching, and cross-platform packaging.
- Full-stack web and mobile development with React, React Native, Expo, Next.js, TypeScript, and Node.js
- Serverless architecture with AWS Amplify Gen 2, Cognito, DynamoDB, S3, Lambda, and GraphQL
- AI workflows and developer tooling with Gemini, MCP, and agent orchestration
- Cross-platform desktop applications with Electron, Vite, and local-first data handling
- Product UX with Tailwind CSS, shadcn/ui, Radix UI, and accessible component systems
I value reliable systems, clear product thinking, and practical automation. My public repositories show the tools and products I actively build.