Evidence before spend
A candidate must retain a dated source, supplier context and economic assumptions before it can enter review.
Case study · active build
ZeroStock Commerce OS
A private operating system for researching online-commerce opportunities, applying guardrails, and learning through small, controlled tests before any inventory or scale decision.
The problem
Opportunity research tends to turn into scattered tabs, intuition and irreversible decisions.
ZeroStock turns that work into a traceable sequence: record the evidence, calculate conservative unit economics, apply a defined policy, request an owner decision and register a bounded test. The system is deliberately private: it is not a customer SaaS product and it does not expose an operator dashboard to the public.
Rules before score
Deterministic margin, delivery, return and loss limits can block a promising-looking candidate.
Human approval at the boundary
A recommendation can create only an internal test record; it cannot publish, buy, charge or contact anyone.
Current scope
Building the operating core before connecting external execution.
- Private owner authentication and role protection
- Persistent candidate, recommendation, decision and test-run records
- Versioned policies for budget, loss, margin, delivery and returns
- Public offer pages only after a verified offer is explicitly published
Engineering approach
Automation should make judgment faster — not hide accountability.
This project reflects how I approach systems with operational and financial consequences: clear boundaries, durable records and human control at the point of commitment.
Connect on LinkedIn