AI skills and review tools for real-estate operators
- Client
- MetaProp Labs
- Period
- Jan 2026 – present
- Role
- External contractor
Problem
Each real-estate operator has its own documents and workflows: rent rolls, T12s, broker emails, offering memos. MetaProp's platform runs client-specific skills; they needed an engineer who could ship reliable, cited skills fast and plug them into reviewers' tools.
What I built
- Trigger.dev skills that parse rent rolls and T12 statements and produce a rent-study asking range, with browser-automation sessions for data collection.
- Deal screener: broker email → structured extraction → reviewer app, in production since Sep 2026. Shipped the MCP endpoint that returns the reviewer's human-corrected values to downstream agents.
- Every skill follows one contract: versioned output envelope, citations, methodology, cost telemetry, honest gaps; synced into the platform through MetaProp's task-runtime SDK.
Result
Client skills that run unattended and produce auditable outputs. Reviewer corrections flow back to agents instead of being lost.
Stack
- Next.js 15/16
- TypeScript
- Trigger.dev
- Drizzle
- Postgres
- Azure AI Foundry
- DBOS
- Entra SSO
- MCP
- Browserbase