> research.divisions
Research & Development Divisions
Verex operates eight core R&D programmes spanning AI-driven detection, computational fire engineering, cyber-physical security, and autonomous suppression. Each division advances the frontier of fire science with rigorous, peer-reviewed research.
> ls --divisions
Eight Disciplines. One Mission.
Each division operates at the intersection of scientific research, advanced engineering, and real-world fire safety challenges.
> cat active_programmes.log
Active Research Programmes
Our laboratories are currently advancing these frontier research initiatives, spanning probabilistic modelling, neuromorphic hardware, autonomous systems, and privacy-preserving machine learning.
Stochastic Fire Growth Modelling Using Bayesian Neural Networks
Probabilistic fire spread prediction combining Monte Carlo sampling with deep learning, applied to compartment fire dynamics under uncertain ventilation conditions. Our Bayesian neural network architecture quantifies epistemic uncertainty in heat release rate predictions, enabling probabilistic tenability envelopes that far surpass deterministic FDS outputs.
Autonomous Multi-Agent Suppression Orchestration
Distributed AI systems where individual suppression nodes communicate via low-latency mesh networks to coordinate optimal suppression response without centralised control. Each node runs a lightweight reinforcement learning policy trained on thousands of FDS-simulated fire scenarios.
Quantum-Resistant Cryptography for Fire-Life-Safety Networks
Post-quantum lattice-based encryption schemes for BACnet Secure Connect, protecting critical fire-life-safety infrastructure from future quantum computing threats. Implementing CRYSTALS-Kyber key encapsulation and CRYSTALS-Dilithium digital signatures within constrained embedded fire panel environments.
Neuromorphic Sensor Arrays for Sub-Second Fire Detection
Spiking neural network hardware deployed on Intel Loihi 2 neuromorphic chips for microsecond-latency fire signature recognition in high-risk industrial environments. Event-driven computation eliminates the polling overhead of conventional detectors, achieving sub-millisecond classification at under 1 mW power draw.
Physics-Informed Neural Networks for Real-Time Structural Fire Response
PINNs that embed Navier-Stokes and heat transfer PDEs directly into neural network loss functions, enabling real-time structural integrity prediction during active fire events. The model ingests live thermocouple and strain gauge telemetry to forecast steel yield and concrete spalling thresholds.
Federated Learning for Privacy-Preserving Fire Risk Intelligence
Cross-organisational fire incident data sharing using federated ML that keeps raw data on-premise while aggregating encrypted model updates via secure aggregation protocols. Building owners and fire brigades contribute to a superior risk prediction model without exposing sensitive tenancy or incident data.
> init --collaboration
Collaborate With Our Research Team
Whether you need computational fire modelling, AI detection research, or cyber-security assessment for critical fire infrastructure — our scientists and engineers are ready to collaborate.
Initiate Research Partnership