WYRMData
Connect local AI workloads to the infrastructure that supports them. WYRM Data explores compute, rack, power and cooling choices using defined engineering checks and traceable asset evidence, from a personal appliance to a shared office or site deployment.
Workload requirements, engineering checks and site evidence in one infrastructure concept.
What WYRM Data is
From workload spec to installed rack. One workspace.
The concept connects workload requirements to candidate compute, power and cooling arrangements, with engineering evidence behind each choice.
Workload in. Spec out.
Workload, user and site requirements form a structured brief for compute, memory, networking and power.
Algorithmic rack recommender
A rule-based approach connects hardware choices, network capacity, cooling and rack space with a traceable bill of materials.
Site and supplier evidence
Compare approved site and supplier records against power, space, cooling and support requirements. Use customer-supplied quotations; any supplier contact needs separate authorisation.
3D rack context
Rack views connect equipment positions with component records, power requirements and configuration notes.
Interactive example
A data hall in context.
A scale model of a working cold-aisle: server cabinets, top-of-rack switches, overhead fibre tray with traffic moving through it, CRAC unit, raised-floor grilles with cooled air rising into the racks. Click anything to inspect its sizing, throughput, thermal envelope, and the standards it's bound to.
drag to orbit · use + / − to zoom
Illustrative CFD airflow + thermal modelling
A locked CFD screening run, rendered live.
An interactive airflow illustration for a representative 42U / 30 kW GPU rack. Explore supply air, warming through the equipment, exhaust and recirculation using a fixed set of inputs. The view is an indicative design-stage screen.
Max rack inlet
28.8 °C
Rack ΔT
15.6 °C
Recirculation
44 %
Air provisioning
0.71×
ASHRAE allowable A1 (≤32 °C) · k-ω SST (low-Re, near-wall resolved) · 1,008,126 cells, 600 iterations. Locked, indicative design-stage screening over a representative 42U / 30 kW rack — not a certified solve.
How it works
Four steps; the loop closes on the fourth.
An illustrative workflow connects a workload brief, equipment configuration, site records and example quotations.
1. Define the workload
A structured brief captures workload, memory, networking, redundancy, location and budget.
2. Compare configurations
Hardware options are considered together: accelerators, chassis, memory, network, cooling and total power.
3. Assess the site
Site records provide the context for power density, connectivity, cooling, location and support requirements.
4. Review the options
Compare approved quotations and technical departures in a draft decision pack. Procurement teams retain authority over supplier contact and orders.
Operator · data.wyrm.ai
exampleBuild · spec · region · quotes
Rack · 42 U · front
0 / 9 components placed
Specification
16× H100
SXM5 80 GB · NVLink 4.0
14.2 kW
one rack · air cooled headroom 6 kW
£7,840 /mo
colo + power + remote-hands + bandwidth
£382k
hardware + install + cabling · 36-mo amort
200 G IB
non-blocking · 25 G Ethernet front-side
100 TB NVMe
Pure //S · 12 GB/s read sustained
Example Site A
United Kingdom · LON-N
- Tier III · 50 kW/rack capacity (you draw 14.2 kW)
- Assurance records: example review fields
- Sub-2 ms to LINX, LON1, LON2 internet exchanges
- Room to grow to 4 racks within the same suite
Example Site A
PendingHarlow
Power-included tariff · 30 A circuits
Example Site B
PendingSlough
ECX Fabric · cross-connect bundle
Example Site C
PendingDocklands
Carrier-dense · 280+ networks on-site
Who WYRM Data is for
Four perspectives on infrastructure planning
For teams connecting AI workloads with the physical infrastructure and engineering decisions behind them.
Local AI deployment teams
Assess the hardware envelope for customer-controlled inference, with measured memory, power and concurrent workload limits.
ML and data-science leads
Right-size training and inference capacity without overbuying. Communicate a spec to vendors that doesn't change shape every meeting.
MEP and data-centre design consultancies
Explore how rack specifications could connect to clearance-aware layouts and the engineering asset record.
Infrastructure acquisition teams
Compare approved hardware and support options with dated quotations, explicit assumptions and technical review.
Engineering extensions
When the rack design becomes a drawing set.
Infrastructure planning reaches across layouts, CAD models, airflow studies and energy assessment. These are connected areas of the Data concept.
Layout planner with manufacturer clearances
PlanningAlgorithmic 2D and 3D layout from the same rule discipline as the recommender. Clearance rules, manufacturer family libraries, hot/cold aisle topology.
Revit + AutoCAD Interface
PlanningModel exchange links rack geometry and clearance envelopes with the wider building design.
Revit plantroom populator
EngineeringExplore plantroom layouts linked to verified rack loads and cooling requirements. Equipment duties and manufacturer constraints require engineering review.
Constrained pipework routing
EngineeringInvestigate routing with defined clearances and clash checks against an approved model. Incomplete geometry and unresolved constraints remain visible to engineers.
CFD airflow + thermal modelling
EngineeringAirflow and thermal studies connect rack-inlet temperatures, containment and recirculation with cooling-design decisions.
Design-stage energy + PUE
EngineeringScreen the design's energy and PUE envelope and produce the Design Stage EPC / SBEM build-up for the building shell. Design-stage screening, cited and versioned — not a lodged certificate.
Shared asset records and defined engineering checks connect infrastructure planning with the wider WYRM approach.
Roadmap
From rack requirements to building design
Rack planning sits within a wider engineering workflow: coordinated models, specifications, schedules and design assessments.
Design
connect equipment choices with the building systems that support them.
The destination
An advanced RIBA Stage 3 MEP design package.
Models, schedules and specifications linked through a consistent engineering record.
Connected engineering outputs
Revit models — real geometry
Equipment geometry, connectors and parameters aligned with manufacturer records and coordination needs.
Specifications
NBS-aligned M&E specifications generated from the design data — kept in lock-step with the model as it changes.
Schedules
Equipment, valve and load schedules linked to the design model and its revision.
TM54
CIBSE TM54 operational energy prediction from the populated systems — modelled in-use performance, not just compliance.
TM65
CIBSE TM65 embodied-carbon assessment of the M&E plant, computed from the scheduled equipment.
BRUKL
Design records supporting Part L assessment and the associated BRUKL submission process.
An example: the specification it writes
Each tab illustrates a piece of M&E plant and its specification context, including power, cooling strategy and fault levels.
crac-chiller-spec-r2.docx
Ready to export
CRAC & chilled-water plant specification
Hall 2 — N+1 chilled-water cooling · 1.20 MW IT load · ASHRAE A1
Illustrative mock. Each item is populated from the same rule discipline as the rack recommender — sized to your design's kW, cooling strategy and fault levels. Switch tabs to flip between the M&E plant a release writes into the Stage 3 specification.
A consistent engineering record connects the workload brief, physical equipment and building systems.