AVIATION SUPPLY CHAIN

On-premise, high-availability AI platform design for an aviation network

A regulated aviation network needed AI agents that run entirely inside its own data center. Wexa delivered the full technical architecture and on-prem deployment design, from microservices to GPU nodes.

Download this case as PDF Book a demo

on-premise: data, models and audit inside the boundary
H100 GPU nodes in the production HA design
control-plane and worker nodes, load-balanced
architecture documentation (arc42 style)

CLIENT: B2B procurement network connecting airlines and suppliers
REGION: US / global
WHEN: 2025
ENGAGEMENT: Solution design
WEXA COMPONENTS: Full Wexa platform · on-prem Kubernetes · GPU inference · 200+ connectors

01Case 42

The challenge

Airline and supplier data cannot leave the network
Agents must integrate with CRM, email, messaging and document stores
Production needs high availability, observability and a test strategy

What Wexa built

Microservices: web, identity, data, workflow and schedule workers, event service, knowledge base, LLM gateway
Durable agent execution on Temporal; MongoDB, Qdrant vector store, Redis, object storage and vault-managed secrets
Kubernetes with Helm, NGINX ingress, autoscaling; Prometheus, Grafana, Alertmanager and Loki
Local GPU inference alongside hosted models through one gateway
Production HA: HAProxy and Keepalived virtual IP, three control-plane nodes, five workers, shared SAN/NAS

02Case 42

How it runs on Wexa

FigureHow this use case runs on Wexa: client systems connected, ontology discovered, context graph in the Wexa memory layer, agents and process flows, and governance. Highlighted components are central to this use case.

03Case 42

Technical architecture

04Case 42

On-prem HA deployment

FigureProduction on-prem high-availability deployment topology.

05Case 42

Results

Measure	Before / without Wexa	With Wexa
Procurement-support staff across the network	40 staff	24 staff
Annual cost of the process	$2.8M	$1.7M
Time to implement the AI solution	9 months	12 weeks
AI build cost (one-time)	$1.2M	$133K
AI operating cost per year (models, infrastructure, upkeep, platform)	$704K	$336K
Total AI cost in year 1	$1.9M	$469K