Defense · AI-powered maintenance
MVC FieldGuide
Your maintenance expert always in the field.
MVC FieldGuide reads your S1000D technical manuals for you: an AI-powered RAG assistant that gives operators step-by-step answers from the approved procedure.
Technology Partner
HPE ProLiant Compute EL2000
−40 to 55 °C
Ruggedized RAG
AI that answers only from your approved S1000D data, on an HPE ProLiant EL2000 rated for −40 to 55 °C, MIL-STD-810H and 461G.
Offline
Deployable at the tactical edge
Fully offline on one rugged 2U box. No network, no cloud, no reach-back: model, index and data stay on the server.
0 GPUs
CPU-only
CompactifAI makes models 60–80% smaller, so FieldGuide runs on a single Intel Xeon 6.
What the operator gets
Any operator, any skill level
- Guided, step-by-step help outside their specialty
- Direct answers for experienced technicians
- Plain-language questions, no query syntax
Traceable to the data module
- Answers come only from approved technical data
- Every answer cites its source DMC
- The operator can open the source and check
Built for disconnected ops
- No network needed at the point of repair
- Model, index and data stay on the server
- One rugged 2U box, no GPU
How it works
Everything runs on the Box
S1000D data in, cited repair procedure out, all on one box
S1000D technical data
Data modules from your CSDB: procedures, descriptions, fault isolation, parts data
Structured ingestion
Indexed by data module, keeping the DMC and document structure intact
FieldGuide engine
Retrieval + CompactifAI compressed LLM running on the Xeon 6 CPU
Answer to the operator
Step-by-step instructions with the source DMC to open and verify
Your team runs the ingestion on site. Multiverse Computing never needs access to your confidential documentation.
CPU-only by design
No GPU required. FieldGuide runs CPU-only.
Compression is what lets a capable LLM run on a standard server CPU, and that changes what you have to deploy, power and support in the field.
Compression makes it fit
CompactifAI shrinks models by 60–80% with 2–4% accuracy loss, so they run from server memory on the CPU.
AI acceleration in the CPU
Intel AMX matrix engines are built into every Xeon 6 core and speed up LLM inference.
Lower size, weight and power
No accelerator card: one less high-power component in a vehicle or shelter.
Simpler to field and sustain
Fewer parts to qualify, cool and keep in spares. The chassis’ second sled slot stays free.
Getting Started
Pilot
FieldGuide on one platform’s S1000D data set, on the EL2000 reference configuration.
Validate
Accuracy checked against a query set agreed with your maintenance team at kickoff.
Field
Roll out to units, with ongoing support and updates as your technical data changes.
Why Multiverse Computing
25
RAG engagements delivered or in delivery
14
Of those, full delivery rather than proofs of concept
Delivered across cloud, on-premise, edge and embedded environments in 7 countries. Edge and sovereign deployment are proven capabilities, not roadmap promises.
Talk to us about a pilot on your own S1000D data.
Pilot FieldGuide on one platform’s S1000D data set, on the EL2000 reference configuration.