Whiteboard — ML-Powered Real Estate Operations Platform

A business operations platform built for a real estate transaction coordination company. Beyond standard transaction tracking and billing, the standout feature is a suite of 20+ ML-powered signal detectors that surface actionable business intelligence in real time.

Why the ML layer matters: Transaction coordination companies track dozens of agents, each with their own submission patterns, revenue contributions, and performance trends. Manual monitoring doesn’t scale. The signal detection system establishes per-agent baselines using historical data, then flags deviations automatically.

Signal detectors include:

• Hot/cold streak detection — identifies agents whose performance is trending significantly up or down relative to their personal baseline
• Cadence anomaly detection — flags when an agent’s contract submission frequency deviates from their established pattern (early warning for churn or capacity issues)
• Revenue forecasting — projects future revenue based on current pipeline and historical close rates
• Concentration risk analysis — alerts when too much revenue depends on a single agent or client
• Year-over-year comparisons — contextualizes current performance against seasonal patterns

All 20+ detectors are built with php-ml, establishing individual agent baselines rather than comparing against company averages — because an agent who normally submits 3 contracts/month dropping to 1 is a very different signal than a high-volume agent dropping from 30 to 28.

Architecture: Symfony 7 backend with Doctrine ORM and FOSRestBundle, React 18 + TanStack Query frontend (with some legacy Vue 2 still in place — a real-world codebase, not a greenfield rewrite). Pushover integration for real-time alert notifications when signals fire.