7 services · from pilot to production

AI that runs where
your business happens.

Shelf intelligence for retail, vision for the factory floor and private LLMs for regulated teams — shipped as products, APIs and engineering you can buy today.

23ms

LLM inference on one Jetson AGX Orin

111×

Speedup from NeuraTensor CUDA kernels

24.83FPS

Real-time vision, concurrent with the LLM

0B

Data egress in air-gapped deployments

What we offer

Pick a starting point.

6 services on one stack — Neuravision, Caroline and the NeuraTensor engine — plus Kilby, our cloud for AI apps.

neuravision · shelf scan · sample
LiveAI shelf auditing for retail chains

Neuravision Retail

Store teams photograph a shelf; Neuravision measures it against the active planogram and flags gaps, misplaced products and compliance issues. Head office gets live dashboards by store, category and team.

  • Shelf scans from up to three photos per gondola
  • Versioned planograms, products, prices and store catalog
  • Alerts with manual review, trends, coverage and exportable reports
  • Role-based access for analysts, replenishers and administrators
  • Bulk Excel import with validation and templates
POST /api/points
$ curl -X POST \
  -H "Authorization: Bearer nv_live_…" \
  -F images=@shelf.jpg

{
  "status": "ok",
  "findings": [ … ],
  "usage": { "calls": 1 }
}
LiveShelf and product image analysis over REST

Neuravision API

Plug the same vision models that power Neuravision Retail into your own apps, ERP or field tools. Self-serve API keys and usage tracking from the developer console.

  • REST endpoints for image analysis and shelf measurement
  • Self-serve API keys with per-key usage and cost tracking
  • OpenAPI documentation and copy-paste quickstart
line-03 · inspection
pass
pass
defect
pass
24.83 FPS · on-device
PilotOn-device visual quality control

Factory & Industrial Inspection

Real-time inspection that runs next to the production line, not in the cloud. Neuravision detects defects, missing parts and anomalies while Caroline explains and logs them — on a single NVIDIA Jetson AGX Orin.

  • Real-time detection at 24.83 FPS (YOLOv8n) on Jetson AGX Orin
  • Vision and LLM reasoning run concurrently on the same device
  • No data leaves the plant: fully air-gapped operation
perimeter · air-gapped
egress 0 B
Caroline · 23 ms
AvailableSovereign LLMs inside your perimeter

Private & On-Prem AI

Deploy NeuratronLLM-Edge 4B (Caroline) where cloud AI is not an option: defense, healthcare, finance and regulated industry. We install, tune and support it on your hardware.

  • 23 ms inference on a single NVIDIA Jetson AGX Orin
  • No cloud dependency, no telemetry, no outbound calls
  • Adaptation to your domain documents and workflows
profile · before / after
baseline1×
neuratensor111×

latency (lower is better)

AvailableFaster models on embedded GPUs

Edge AI Engineering

Our team ports and optimizes your models for NVIDIA Jetson and Ampere-class GPUs using the NeuraTensor stack — the same custom CUDA kernels behind our 111× speedup.

  • Profiling and bottleneck analysis of your current pipeline
  • Custom CUDA and fused SNN-SSM kernels via NeuraTensor SDK
  • Memory reduction and latency budgets for real-time systems
AvailableTry the full stack on your own use case

Evaluation & Pilot Program

A fixed-scope evaluation to validate Neuramorphic on your data before committing: we define success metrics together, run the stack on a Jetson AGX Orin and deliver a results report.

  1. 01

    Success criteria agreed up front

  2. 02

    Access to Caroline, Neuravision and the NeuraTensor SDK

  3. 03

    Benchmarks on your data and hardware

  4. 04

    Results report with a production roadmap

LiveCloud platform · onkilby.com

The cloud for AI apps, agents and models.

Kilby builds and runs your AI apps, agents and model servers on AWS, with logs, rollbacks and your own domains included.

  • Git, image or template
  • Postgres with pgvector
  • An MCP server for agents
  • Private services
onkilby.com/dashboardIllustrative
AI workspace3 services · 1 database
  • open-webuiWeb serviceLive
  • ollamaPrivate serviceLive
  • mainPostgres · pgvectorAvailable
  • agent-apiWeb service · from GitDeploying
terminal
$ kilby deploy --repo acme/agents
# or let your coding agent do it
$ kilby mcp

How we work

Weeks to value, not quarters.

  1. 01

    Discover

    A working session on your stores, plant or data constraints. We agree on success metrics before anything is built.

  2. 02

    Pilot

    Fixed scope, your data, real hardware. You get benchmarks and a results report — not a slide deck.

  3. 03

    Production

    Roll-out across sites with support, monitoring and model updates. API, on-prem or fully managed.

Not sure which one fits?

Tell us about your stores, plant or data constraints. We'll come back with the shortest path to production.

Book a call