Cogentic
SECURITY & DEPLOYMENT

Your data never leaves your environment.

Cogentic installs on a workstation or server inside your network and runs the whole AutoML pipeline there: no cloud account, no uploads, no outbound calls. Updates arrive as offline packages your team applies inside the boundary. Models, logs, and audit trails stay on disk, on hardware you own. That is how air-gapped machine learning runs in a plant that can't send data out.

On-premAir-gap capable
DEPLOYMENT ARCHITECTURE

How we deploy in your plant.

Three topologies, one invariant: everything runs inside a boundary you own. Start on a single workstation and move to a server or a fully air-gapped install without changing the guarantee.

Your network
Historianprocess data
MES / SCADAline data
CSV / Excelexports
Cogentic workstationprofile · train · predict
Models & predictionsstay on your hardware
0 outbound connections
Verifiable with a packet capture
01

Desktop install

One workstation
CSV · Excel · Parquetyour files
Cogentic appprofile · train · predict
Models on disklocal project folder

Installed on a single engineer's machine. Data, models, and results live on that workstation. Nothing else to stand up.

  • Fastest way to a pilot
  • No server, no network dependency
02

On-prem server

Your network
Engineer workstationsyour team
Cogentic servershared projects
Models & logson your server

Deployed on a server inside your network so a team shares projects and models, tailored to your operation, all within your perimeter.

  • Multi-seat, one boundary
  • Nothing leaves your perimeter
03

Fully air-gapped

Isolated network
Offline installerdelivered as media
Cogentic workstationprofile · train · predict
Models & audit trailon disk, inside
0 outbound

Installed from offline media on machines with no outbound connectivity, for environments governed by your ITAR or CMMC controls.

  • Zero outbound connectivity
  • Offline update packages
Updates

Updates ship as offline packages: a code-signed, versioned installer with a SHA-256 checksum your team verifies and applies inside the boundary, on your schedule. Feature releases land quarterly; critical fixes ship as needed. No auto-update, no phone-home, and the previous version stays available so you can roll back. Air-gapped sites receive the same signed package on media.

What lives where
Datasets
On disk in your project workspace, on the machine you installed on. Never copied offsite.
Trained models
Saved as files inside the boundary. Portable on your terms: you can back them up, move them, or delete them.
Run logs & audit trails
Written locally alongside each run and exportable for your QA records. No telemetry leaves the machine.
Cleaning & training plans
Stored as hashed, deterministic contracts next to the models they produced, so every result stays reproducible.
Prerequisites
Operating system
Windows 10/11, macOS 12+ (Apple silicon or Intel), or Ubuntu 20.04+ / RHEL 8+, all 64-bit.
Hardware
A standard engineering workstation: 4+ cores, 8 GB RAM (16 GB recommended for multi-GB datasets). No GPU required. Training runs on CPU; NVIDIA acceleration is optional for very large datasets.
Data formats
CSV, Excel (.xlsx), and Parquet exports from your existing systems.
Network
None required. Installation and operation work fully offline.
Timeline

A desktop pilot is typically running on your data within the first week. An on-prem or air-gapped install typically lands in two to four weeks, including tailoring to your operation and working through your security review.

Who does what

We handle

  • Building the offline installer and update packages
  • Tailoring the workflow to your data and operation
  • Documentation for your security review
  • Training your engineers on upload → profile → train → predict

Your team handles

  • Provisioning the workstation or server
  • Applying installers and updates inside the boundary
  • Network, accounts, and access control, as you do today
  • The data itself. It never passes through us.
COMPLIANCE-READY

Fits within the boundaries you already operate.

No new vendor cloud to trust. Cogentic deploys inside your own controls and never sends data out, so it works within the frameworks you already answer to, such as SOC 2, HIPAA, or GDPR.

A cloud tool turns every prediction into a data-transfer decision someone has to sign off. Keeping the work in-house removes it, and the failed review, leaked dataset, or exposed IP that can come with it.

SOC 2HIPAAGDPRCMMCITAR

Deploys inside your existing controls

Installs in your own environment and inherits the boundary you already operate. No new one to stand up.

Data never crosses a boundary you don't own

Nothing uploaded, no outbound calls. No third-party processor to assess, no vendor cloud to clear through review.

Air-gapped installs supported

Runs fully air-gapped on isolated machines, wherever data residency is non-negotiable.

We support your compliance process

Our team works through security review and documentation alongside yours.

SOC 2, HIPAA, GDPR, CMMC, and ITAR are examples, not the whole list. Cogentic is deployable within the controls you already operate. This describes how the software fits your environment, not a claim that Cogentic itself holds any certification.

FOR YOUR SECURITY REVIEW

The paperwork your reviewers will ask for.

Available on request

Security whitepaper

Architecture, data flow, and the zero-egress boundary, written for your security and IT teams.

Available on request

Security questionnaire pack

Standard answers for vendor-review questionnaires, ready to hand to procurement and shorten your review cycle.

SEE IT IN YOUR ENVIRONMENT

See it run inside your environment.

We'll walk through how Cogentic deploys inside your controls, whether on-prem, offline, or fully air-gapped.

Book a demo

Running today on an air-gapped network at a Fortune 100 manufacturer