// AIEdge.jsx — "AI at the Edge" platform deep-dive. // Route: #/platform/ai-edge function AIEdgePage({ onNavigate }) { const [activeArea, setActiveArea] = React.useState(null); React.useEffect(() => { if (!activeArea) return; const onKey = (e) => { if (e.key === "Escape") setActiveArea(null); }; window.addEventListener("keydown", onKey); const prev = document.body.style.overflow; document.body.style.overflow = "hidden"; return () => { window.removeEventListener("keydown", onKey); document.body.style.overflow = prev; }; }, [activeArea]); // 11 ARMS functional focus areas const armsAreas = [ { t: "Aviation Safety Program", b: "Command safety support, hazard reporting, and mishap prevention.", more: "The inference engine cross-checks the unit's safety posture against the current command safety program requirements — hazard reporting workflows, mishap-prevention measures, and the documentation an evaluator expects to see — flagging gaps before they become findings." }, { t: "Aircraft Maintenance", b: "Standardization, production control, and repair / recovery operations.", more: "Trained on the maintenance standardization and production-control criteria, the model reviews the unit's maintenance program for the conditions ARMS evaluators assess — from production control board discipline to repair and recovery procedures." }, { t: "Flight Operations", b: "Aircrew training, currency, and flight scheduling.", more: "Reads the unit's training and currency picture against regulation and surfaces aircrew approaching non-currency, training gaps, and scheduling conflicts that would draw scrutiny during a survey." }, { t: "Supply & Logistics", b: "Command Supply Discipline Program (CSDP), tool room procedures, hazardous material management.", more: "Evaluates the unit against CSDP requirements, tool room control procedures, and hazardous material handling and storage standards — the high-frequency finding areas in any logistics review." }, { t: "Quality Control", b: "Technical manual compliance and component tracking.", more: "Checks technical manual currency and compliance and validates component tracking against the records an evaluator will pull, surfacing discrepancies before the inspection does." }, { t: "Physical Security", b: "Arms, ammunition, and explosive (AAE) storage regulations.", more: "Assesses arms, ammunition, and explosives storage and accountability against the governing physical security regulations — one of the most consequential and tightly-scored survey areas." }, { t: "Petroleum, Oils & Lubricants", b: "Spill response plans and FARP operations.", more: "Reviews POL handling, spill response planning, and Forward Arming and Refueling Point (FARP) operations against environmental and operational standards." }, { t: "Avionics & Calibration", b: "TMDE compliance and system inspections.", more: "Validates Test, Measurement, and Diagnostic Equipment (TMDE) calibration currency and avionics system inspection compliance against the calibration cycle the survey enforces." }, { t: "Fire & Crash Rescue", b: "Alarm system tests and pre-accident plans.", more: "Confirms alarm system test records and the currency and adequacy of the unit's pre-accident plan against the standards evaluators verify." }, { t: "Standardized Checklists", b: "Aircrew read-do / flow-verify sequences across all mission flight phases.", more: "Compares the unit's aircrew checklists — read-do and flow-verify sequences across every mission phase — against the approved standardized references, flagging deviations and out-of-date procedures." }, { t: "Weight & Balance Program", b: "Verification of aircraft weighing procedures and documentation.", more: "Audits aircraft weighing procedures and weight-and-balance documentation for the currency and completeness an ARMS evaluator requires." }, ]; const compare = [ ["Where it runs", "Embedded in the ASD application tier", "On dedicated edge inference units, alongside any system"], ["Scope of knowledge", "ASD's own data, workflows, and records", "Regulations, standards, and inspection criteria, force-wide"], ["Primary job", "Summarize and analyze the unit's live operating data", "Assess the unit against approved doctrine and criteria"], ["Connectivity", "Runs wherever ASD runs, including disconnected", "Disconnected by design — inference never leaves the edge"], ["Output", "Insight on what the unit is doing now", "Findings on where the unit stands against the standard"], ]; return (
{/* ─────── breadcrumb / sticky meta ─────── */}
{ e.preventDefault(); onNavigate("/"); }}>Index / { e.preventDefault(); onNavigate("/platform/aviation"); }}>Platform / AI at the Edge
DOC · LI-PF-003 · REV 1
{/* ─────── hero ─────── */}
PLATFORM AI · INFERENCE AT THE EDGE

AI at
the edge.

LLMs trained on Army regulations, standards, and inspection criteria, deployed on AI-enabled edge processing units. The result is the most sophisticated, knowledgeable, and approved inference engine ever fielded at the unit — analysis that runs where the mission is, not in a cloud the unit can't reach.

{ e.preventDefault(); onNavigate("/contact"); }} className="btn"> Brief us → { e.preventDefault(); onNavigate("/platform/aviation"); }} className="btn ghost"> See ASD
{/* ─────── hero media full-bleed ─────── */}
{/* Spaces in the filename must be URL-encoded (%20) — nginx 404s on raw spaces. Same pattern as the Rapid Prototyping Loop hero. */}
{[ ["DEPLOYMENT", "On-device"], ["MODELS", "Army-trained LLMs"], ["CONNECTIVITY", "Disconnected-capable"], ["USE", "Unit self-assessment"], ].map(([k, v], i) => (
{k}
{v}
))}
{/* ─────── overview ─────── */}
01 / Overview

The standard, on-device. The inspector's knowledge, before the inspection.

Edge processing units running models trained on regulatory text, doctrinal standards, and inspection criteria put an approved inference engine inside the unit — no connectivity required, no sensitive data leaving the perimeter. The unit can ask the regulation directly, and get an answer grounded in the criteria it will actually be measured against.

{/* ─────── how it works ─────── */}
02 / How it works

Trained on the
standard itself.

The models are trained and grounded on the regulations, standards, and inspection criteria that govern Army aviation — not general internet text. Inference runs locally on edge hardware, so the analysis is fast, private, and available in the disconnected conditions units actually operate in. Every answer traces back to an approved source.

{[ ["01", "Ground", "Models trained on regulations, standards, and inspection criteria."], ["02", "Deploy", "Inference runs on AI-enabled edge units at the unit, disconnected-capable."], ["03", "Assess", "The unit queries the standard and assesses itself before the evaluator arrives."], ].map(([n, t, b], i) => (
{n} ——
{t}

{b}

))}
{/* ─────── ARMS self-assessment ─────── */}
03 / ARMS self-assessment
AVIATION RESOURCE MANAGEMENT SURVEY

Assess yourself
before the survey.

The Aviation Resource Management Survey (ARMS) evaluates aviation units and support facilities against FORSCOM and USACRC standards — roughly 600 to 700 items across 11 functional core areas. The official tools live on the U.S. Army Combat Readiness Center and JTDI, behind CAC. Lulius edge models, trained on those same standards, let a unit run the assessment on itself first — finding the gaps while there is still time to fix them.

600–700
Checklist items
11
Functional core areas
2
Standards bodies · FORSCOM · USACRC
{/* 11 functional areas grid */}
{armsAreas.map((a, i) => ( ))}

* Checklist content is tailored to a unit's mission set. Official ARMS tools require CAC access via USACRC / JTDI.

{/* ─────── Platform AI vs ASD AI ─────── */}
04 / Two engines

Platform AI,
and ASD AI.

ASD already has an LLM at the edge built in — it reads the unit's own live data and turns it into insight. Platform AI inference is a different engine: trained on the regulations and inspection criteria themselves, it measures the unit against the standard. The two work together — one tells you what you're doing, the other tells you where you stand.

ASD AI · BUILT IN
PLATFORM AI · AT THE EDGE
{compare.map(([k, asd, plat]) => (
{k}
{asd}
{plat}
))}
{/* ─────── why edge / posture ─────── */}
05 / Why the edge

Approved. Private.
Always available.

Putting inference on the edge is a posture decision, not just a performance one. The model is approved and grounded on authoritative sources, sensitive data never leaves the unit, and the analysis is available in exactly the disconnected, degraded conditions where a cloud service is not. It is the same field discipline that defines every Lulius system — applied to AI.

BRING IT TO YOUR UNIT
See edge inference assess against the standard
{ e.preventDefault(); onNavigate("/contact"); }} className="btn"> Request a briefing →
{/* ─────── ARMS area details dialog ─────── */} {activeArea && (
setActiveArea(null)} style={{ position: "fixed", inset: 0, zIndex: 1000, background: "rgba(0,0,0,.78)", display: "flex", alignItems: "center", justifyContent: "center", padding: 24, backdropFilter: "blur(4px)", WebkitBackdropFilter: "blur(4px)", }} >
e.stopPropagation()} style={{ maxWidth: 720, width: "100%", maxHeight: "85vh", background: "var(--ink)", border: "1px solid var(--accent)", padding: "clamp(28px, 4vw, 48px)", overflowY: "auto", position: "relative", boxShadow: "0 30px 80px rgba(0,0,0,.5)", }} >
ARMS · FUNCTIONAL CORE AREA

{activeArea.t}

{activeArea.b}

{activeArea.more && (

{activeArea.more}

)}
)}
); } window.AIEdgePage = AIEdgePage;