Skip to content

Webinar · recording available

Secured AI: Containerized on IBM Fusion, powered by NeuralSeek.

A thirty-minute session on running generative AI behind your own firewall: IBM Fusion as the containerised foundation, NeuralSeek as the platform on top, and the applications that work on the first day. Recorded September 15, 2026.

What was covered

Four arguments, and a short demonstration.

The session was built around questions a security or compliance lead asks before an AI purchase. These are the answers it gave.

  • Data sovereignty as the deciding question

    The most common objection in the room was not accuracy but jurisdiction: where does the data go, who can be compelled to hand it over, and what happens if a provider withdraws a model. The answer given: a solution that works with the internet unplugged, on models you host, under a retention policy that is yours.

  • IBM Fusion as the foundation

    IBM Fusion is the on-premises, containerised infrastructure: compute, storage, networking and backup on Red Hat OpenShift, in your building. One vendor to call, one environment to prove to an auditor, instead of a mix of spare servers, borrowed GPUs and over-committed virtual machines.

  • NeuralSeek as the layer on top

    The platform that makes the hardware useful on day one: guardrails, auditing and governance built in, a visual view of every agent, and the freedom to swap the language model without re-tuning anything else — because the best model is rarely one model; it depends on the task, the latency and the cost.

  • Applications that work on the first day

    The demonstration showed the bundled applications rather than an empty platform: a meeting listener that keeps transcripts and summaries on your own infrastructure, and the Word, Excel, PowerPoint and Outlook add-ins that run against the models you chose. Shadow AI was the argument for shipping something usable immediately.

How the session ran

The order things were taken in.

Slides first, then the product, then the questions.

  1. Why AI purchases stall

    Hardware without applications, months of building before anything is shown, and the pressure that puts on the team that signed the order.

  2. The IBM Fusion foundation

    What the appliance provides underneath: OpenShift, storage, backup and the operational consistency of one hardware platform.

  3. Sovereignty, retention and provider risk

    Cross-border data transfer, compelled access, cloud-provider retention of queries, and models that have been withdrawn without notice.

  4. Demonstration

    The bundled applications, the model-swap workflow, and the Office add-ins, shown in the product.

  5. Questions

    How long it takes an engineer with no model experience to become productive, and whether email can be used as a data source.

Who it was for

The people who have to sign off.

The session was written for the reviewers, not the builders — the recording is still the shortest way to see the argument made end to end.

  • IT and security leaders in healthcare

    Where patient data cannot leave the perimeter and every AI vendor is another agreement to track.

  • IT and compliance leaders in financial services

    Where retention, examination and cross-border rules decide the architecture before the use case does.

  • CISOs and risk leaders

    Who need one platform to approve once, rather than a new review for every bolt-on.

  • Data and platform engineering teams

    Who will run it, swap the model, and be asked how long until it does something useful.

Presented by

Two voices, one session.

Credited by role. Registration and follow-up were hosted by TechD.

  • Host — partnerships

    NeuralSeek

  • Presenter — solutions

    TechD

See it running inside your own boundary.

Talk to a NeuralSeek expert about how it fits your stack, where your data has to live, and the governance your auditors already expect.