Research Overview

Pioneering integrated thermal architectures for high-density compute. We develop solid-state MHD cooling and liquid-metal thermal loops to breakthrough the >1000W TDP barrier, defining the infrastructural layer for Sustainable AI Factories.

Phase 1: Substrate Simulation NVIDIA-Accelerated AI Workflows Proprietary MHD Flow Models

Active Research Domains

Our current work spans four interconnected areas โ€” each supported by specialised agents in the Aviary fleet. Design decisions are simulation-validated before physical prototyping begins.

โšก

Axial-Flux Motor Topologies

Comparative design, FEA simulation, and optimisation of YASA, TORUS, and coreless axial-flux configurations. Target metrics: torque density >5 Nยทm/kg, efficiency >95% at rated load, power factor >0.9.

Active
๐Ÿงฒ

Halbach Array Engineering

Optimising segment count, orientation, and magnet material for maximum air-gap flux density. Linear and rotary configurations explored. NdFeB N52 and SmCo grades characterised against design requirements.

Active
๐Ÿช

Linear Actuator Systems

High-force-density linear drives using Halbach-magnetised movers and coreless coil arrays. Target applications include precision positioning and high-cycle industrial actuation.

Design Phase
๐Ÿ“

Agent-Driven FEA Automation

All simulation workloads run through the Aviary multi-agent cluster โ€” enabling parametric sweeps over hundreds of design configurations that would be intractable with manual workflows.

Infrastructure
๐Ÿ”ฌ

Core Material Characterisation

Benchmarking soft magnetic composites (SMC), amorphous metals, and grain-oriented silicon steel for high-frequency axial-flux operation. Targeting operating frequencies of 400โ€“2000 Hz.

Research
๐ŸŒก๏ธ

Extreme Heat Flux Mitigation

Modeling complex liquid-metal transfer across 1000W+ TDP surfaces. Designing active MHD cooling pathways for continuous heat extraction at extreme power densities in compact silicon architectures.

Analysis
๐ŸŒ

NVIDIA Omniverse Digital Twins

Building exact physical replicas of our liquid-metal cooling loops within NVIDIA Omniverse. Empowering real-time simulation of fluid dynamics, thermal stress, and failure modes at datacenter scale prior to tape-out.

Simulation

NVIDIA-Accelerated AI Research Integration

MagFlux research is powered by an advanced AI-driven methodology utilizing specialized generative models. Taking advantage of NVIDIA GPU acceleration, our autonomous workflows evaluate complex fluid dynamics and electromagnetic topologies at significantly higher speeds than traditional iterative development. All simulations pass through stringent human-in-the-loop review to ensure viability before physical prototyping begins.

Restricted Vault Access

Full mathematical derivations, research papers, and the interactive parametric design simulator are strictly proprietary and confidential. Access to the Vault requires explicit acknowledgment of confidentiality terms.

Access the Vault Return Home