I’m a software engineer, independent builder, and systematic investor.

I build agent-friendly workflow systems for complex domains — turning domain intent into structured DSLs, DAG/IR pipelines, validation loops, executable tools, and human-facing workspaces.

My current work explores the same pattern across different domains:

• MyInvestPilot: an AI-native investment operating system where strategies are modeled as composable primitives, backtests become observation panels, and decisions can be tracked, explained, and improved.
• CraftDAG / MinePilot: an agent-friendly Minecraft construction compiler and workspace where build intent becomes ComponentPlan, CraftDAG, VoxelPlan, validation, preview, materials, layers, and schematic export.

Human-friendly UI is not enough in the agent era. Software also needs agent-friendly workflows.

• MyInvestPilot An AI-native investment operating system for systematic, rule-based investing.

• Chat2Invest An AI research assistant for structured investment analysis.

• CraftDAG An agent-friendly Minecraft construction compiler: ComponentPlan DSL → CraftDAG IR → VoxelPlan → validation, materials, layers, and schematic export.

• MinePilot An agentic Minecraft build workspace powered by CraftDAG — preview, validate, iterate, and export complex structures and automation plans.

• free4.chat An anonymous WebRTC chat experiment exploring lightweight, real-time social interaction.

• bmpi.dev My long-form writing on software engineering, AI-native products, investing, agent-friendly workflow systems, and independent building.

• i365.tech A community and consulting space around systems, AI, and personal leverage.

• Agent-friendly workflow systems
• AI-native investment systems
• Strategy primitives and rule-based portfolio engines
• Decision intelligence and human-AI workflows
• Domain DSLs, DAG/IR pipelines, validation contracts, and repair loops
• Cloudflare-native serverless architecture
• Independent product building
• Long-form technical writing

• Designing Agent-Friendly Workflows My current architecture thesis: complex AI systems need more than chat boxes. They need domain DSLs, IR/DAG pipelines, validation contracts, repair loops, and workflow surfaces that agents can read, write, validate, observe, and repair.

• Agent 时代的工作流设计 中文版本：通过 MyInvestPilot 和 MinePilot / CraftDAG 两个案例，整理我对 Agent-friendly workflow systems 的理解：Intent → Domain DSL → IR/DAG → Validation → Repair Loop → Execution。

• How I Built an AI-Native Quantitative Investment System My long-form explanation of how I think about AI-native investment systems, strategy primitives, and machine-readable decision infrastructure.

• 我是如何构建一个 AI 原生量化系统的 中文版本：关于 AI 原生量化系统、策略 DSL、规则化投资与个人长期产品构建的完整复盘。

For more essays, notes, and annual reflections, visit bmpi.dev.

Languages
TypeScript · Python · Java · Elixir · Go · Clojure

Infrastructure
Cloudflare Workers · R2 · D1 · Durable Objects · Queues · AWS · Docker · PostgreSQL · Redis

AI / Product
LLMs · Agent-friendly workflows · AI-native systems · Strategy DSLs · Domain DSLs · Automation · SEO · Analytics

• Blog: bmpi.dev
• Product: myinvestpilot.com
• Chat: free4.chat
• GitHub: @madawei2699