AI Suite Enablement for an HVAC BMS Platform

Azendian, a Southeast Asian HVAC and building-systems specialist working with Innoflex, wanted its Building Management System to do more than monitor and control. The goal was to turn the BMS into an intelligent platform that predicts faults, forecasts component health, and trims energy use, capabilities its customers were starting to demand and its competitors did not have.

The platform faced limits common across the HVAC industry:

• Fault detection driven by static rules and manual thresholds
• Reactive or schedule-based maintenance
• Energy optimisation that depended on operator experience
• High false-alarm rates eroding trust in automated alerts
• Customers increasingly demanding measurable energy savings and reliability

VentureSEA designed and deployed a modular AI suite, integrated directly into the client's BMS and proven on live sites.

• Built machine-learning Fault Detection and Diagnosis across multiple live HVAC sites, reaching a 93 percent true positive rate at roughly 7 percent false positives
• Developed Remaining-Useful-Life models to estimate component health and enable predictive maintenance
• Designed load-forecasting and data-driven optimisation to control HVAC dynamically within comfort constraints
• Streamed sensor telemetry through Kafka, trained gradient-boosted models with scikit-learn and XGBoost, tracked every experiment in MLflow, and served models via FastAPI for production reliability
• Shipped a model registry and retraining workflow so recalibrated models can be promoted to live sites without vendor lock-in

Tested across multiple sites, the suite delivered 15 to 20 percent energy savings and repositioned the BMS as an intelligent operations platform.