AI

Atomic Tessellator builds the computational infrastructure for advanced materials development. The platform allows organisations to explore thousands of material candidates through simulation, committing physical resources only to high-confidence options — compressing development cycles from months or years to days.

Headquartered in Auckland, New Zealand, the company targets defence, aerospace, and advanced engineering customers who face extreme material performance requirements. Use cases include new materials for EV batteries, carbon capture, fusion reactors, and replacing substances under regulatory scrutiny (PFAS, microplastics). The platform is built on ab-initio (first principles) methods, GPU-accelerated infrastructure, and AI-driven simulation, deployed via hosted cloud, dedicated cloud, or on-premise.

AI-native autonomous control platform for commercial and industrial buildings — using reinforcement learning and a universal hardware gateway to actively shape electrical demand, cut peak loads by 30%+ and deliver verified energy savings without disrupting operations.

Their insight, forged from direct experience with grid-constrained buildings, is that the energy transition is fundamentally a demand control problem: US commercial facilities waste an estimated $250B annually not because they lack power, but because legacy Building Management Systems are static, rules-based and incapable of adapting to real-time grid conditions. Demand charges alone can represent 30–40% of a facility's electricity bill. Brick's thesis is that an AI-native closed-loop control layer — sitting between existing BMS infrastructure and the grid — can structurally solve this.

The Brick platform has three interlocking components. A universal hardware gateway installs in under ten minutes and connects to any existing BMS or HVAC system regardless of brand — a deliberately brand-agnostic approach that incumbent vendors like Siemens and Schneider are structurally disincentivised to replicate, since their own business models depend on proprietary hardware lock-in. A reinforcement learning model then runs a continuous closed-loop optimisation, shaping electrical load in real time to stay within fixed grid limits while maintaining occupant comfort. ,