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.
Local-first by design, not an enterprise add-on.
Most no-code AutoML tools are cloud services that ship your data offsite. Cogentic runs the whole pipeline on hardware you own.
Runs where your data lives
A desktop app with a local backend. Runs on your workstation or inside your network, never on infrastructure you don't control.
No outbound calls
Nothing uploaded, no account to create. Profiling, cleaning, training, and prediction all happen locally, internet off.
Optional local LLM, or none
Assistive features use a local LLM you host. With none configured, Cogentic falls back to deterministic heuristics. Nothing leaves the machine.
Full offline & air-gap mode
Install once and disconnect. Built for air-gapped machines and isolated networks where outbound traffic is blocked.
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.
Desktop install
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
On-prem 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
Fully air-gapped
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 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.
- 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.
- 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.
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.
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.
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.
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.
Every model traces to an exact, re-runnable plan.
In regulated work, a prediction you can't explain is one you can't use. Deterministic, hashed contracts mean any result can be reproduced and defended.
Hashed, deterministic contracts
Every cleaning and training step is recorded as a hashed contract. Same input, same plan, same result. No hidden state, no drift.
Every model traces to a plan
A model maps back to the exact, re-runnable plan that produced it, from raw data through cleaning to the champion model.
Audit-ready by default
Hand an auditor the plan and the hashes. They re-run it and reproduce the result independently. Built in, not reconstructed.
The paperwork your reviewers will ask for.
Security whitepaper
Architecture, data flow, and the zero-egress boundary, written for your security and IT teams.
Security questionnaire pack
Standard answers for vendor-review questionnaires, ready to hand to procurement and shorten your review cycle.
See it run inside your environment.
We'll walk through how Cogentic deploys inside your controls, whether on-prem, offline, or fully air-gapped.
Running today on an air-gapped network at a Fortune 100 manufacturer