The right people, whenever you need them.

From a single specialist to a complete delivery team – project management, product, architecture, UX, development, QA and DevOps, ready to plug into your business.

Get in touch

A hub, not a silo.

The name says it: a Knoten is a knot – a point where strands meet. Knotenhub is the point where your strategy, systems and people connect.

We work as an independent partner, small and senior, bringing clarity to complex problems and seeing the work through to something that actually ships.

What you can hire us for

Every role a software project needs, from the first plan to running in production – hired as a specialist, or together as one delivery team.

  • Plan and lead

    Someone who owns the outcome, the backlog and the big picture.

    • Project and delivery management

      Scope, budget, risks and stakeholders in one place, and progress reported honestly.

    • Agile coaching and Scrum master

      A team rhythm that fits: Scrum, Kanban or Scrumban, without ceremony for its own sake.

    • Product ownership and business analysis

      A backlog ordered by value, and requirements that developers and testers read the same way.

    • Architecture

      Solution and cloud architecture, with security and compliance designed in and decisions written down.

  • Design and build

    The product itself, clear to use and built to last.

    • UX design

      Research-led design that makes the product clear and easy to use.

    • Development

      The application itself: clean, maintainable software, seen through until it ships.

    • QA

      Testing in the pipeline, not as a phase at the end. The release either holds, or it rolls back.

  • Ship and run

    Into production, and kept there reliably.

    • DevOps and infrastructure

      Into the cloud, between clouds, or back onto your own hardware.

      Migration and modernisation

      Whichever direction pays off for the workload, with the tests to prove nothing broke.

      • Off a PaaS you have outgrown, to a managed platform or your own Kubernetes
      • VMware exit after Broadcom: Proxmox, OpenShift Virtualization, OpenStack
      • Selective cloud exit: the expensive workloads to Hetzner or bare metal
      • AWS, Azure, GCP, or EU sovereign clouds such as STACKIT, OVHcloud and IONOS
      • Containerise the monolith, split only what pays off
      • Landing zones: accounts, IAM, network, done once and properly
      Pipelines and GitOps

      Ship every day without drama. Pipelines your team can read, releases that roll back on their own.

      • CI/CD with GitHub Actions and GitLab CI, Jenkins migrations
      • GitOps with Argo CD or Flux
      • Progressive delivery: canary and blue/green with Argo Rollouts or Flagger
      • Quality gates, integration tests, automatic rollback
      • End-to-end tests with Playwright that run in minutes, not hours
      Infrastructure as code and platform

      Reproducible environments, no snowflake servers, and a platform your developers actually want to use.

      • OpenTofu and Terraform, Pulumi, Ansible, AWS CDK
      • Kubernetes: EKS, AKS, GKE, OpenShift, k3s
      • OpenTelemetry-first observability: Prometheus, Grafana, Loki. Datadog bills cut down to size
      • Secrets and certificates: Vault or OpenBao, External Secrets, automated rotation
      • FinOps: cloud, SaaS and GPU spend, right-sizing, and when self-hosting is cheaper
      AI, cloud or self-hosted

      The same operational discipline applied to models. Hosted APIs where they fit, open-weight models on your own GPUs where volume or data rules justify it.

      • Claude, OpenAI or Mistral via Bedrock, Vertex or EU hosting, or open models on your hardware
      • An LLM gateway for routing, cost tracking and failover
      • Agent workflows and MCP integrations, retrieval, evals, guardrails
      • GPU platform on Kubernetes, inference with vLLM
      • EU data residency, GDPR, and AI Act-ready logging and oversight

Work we have done

Our clients rarely want to be named. Here is what the work looked like.

  • Recruitment

    Self-hosted AI candidate data never left the company

    AI that takes people by the hand

    Problem
    Candidates were matched by hand, and every customer got the same onboarding.
    What we did
    Automatic candidate matching, and guided onboarding that adapts each step to the customer – on a locally hosted model, for data protection.
    Result
    Matches arrive on their own, and customers are led through onboarding step by step.
  • Consultancy

    Ask the wiki not the colleague next door

    An assistant that knows what is written down

    Problem
    Staff asked colleagues for answers that were already in Confluence and the other wikis.
    What we did
    An internal AI assistant that answers questions from the company's own wikis.
    Result
    People find answers themselves, and colleagues are interrupted less.
  • Insurance

    0 manual certificate renewals

    Two clouds, one network, no expiring certificates

    Problem
    AWS and Azure regions had to talk to each other securely, and certificates were renewed by hand.
    What we did
    Connected the regions and automated how certificates are issued and rotated.
    Result
    Certificates rotate on their own. Nobody renews one by hand any more.
  • Media

    Auto rollback on a bad release

    Releases that roll themselves back

    Problem
    Every release to a live audience was a risk, so releases were rare and large.
    What we did
    Progressive delivery: canary releases watched by metrics, with automatic rollback.
    Result
    A bad release stops itself before most users see it.
  • Automotive

    −⅓ cloud costs, about a third

    A multi-cloud bill cut by about a third

    Problem
    Spend across several clouds kept growing, and no one could say why.
    What we did
    Attributed every cost to an owner, right-sized, and removed what nobody used. Instances now scale with demand, with spot instances wherever a workload can handle interruption.
    Result
    Cloud costs down by about a third.
  • Software

    Hours to minutes end-to-end test run

    A test suite from hours to minutes

    Problem
    The end-to-end tests took hours, so they ran rarely and caught problems late.
    What we did
    Rebuilt the suite to run in parallel inside the pipeline.
    Result
    The suite runs in minutes, on every change.

How an engagement works

What happens after you email us.

  1. Intro call

    Thirty minutes. You describe the problem, and we tell you honestly whether we are the right fit.

  2. Assessment

    A short look at your setup, then a written recommendation with options, effort and cost. It is yours to keep, whatever you decide.

  3. Delivery or handover

    We do the work alongside your team, or hand over a plan your team can run with. Either way, it is documented.

Practical questions

Can you provide a whole team?

Yes. From one specialist to a complete team – project management, agile coach, product, architecture, UX, development, QA and DevOps – working as one unit.

Do you work alongside our own people?

Yes, and that is the usual case. We join your tools, rituals and chat, and leave the knowledge behind when we go.

Do you work remotely or on site?

Mostly remotely, from Ireland. We come on site for kick-offs, workshops and whenever the work needs it, in Ireland, the UK and the EU.

Which time zone do you work in?

Irish time. That overlaps fully with UK hours and with most of the working day in Central Europe.

Which languages do you work in?

English and German.

How does pricing work?

A day rate for open-ended work, or a fixed price for a clearly scoped piece. The assessment shows which one fits.

Is there a minimum engagement?

No formal minimum. A short review or a single assessment is a perfectly good place to start.

How fast can you start?

Usually within two weeks. Small, urgent pieces often sooner.

Let's talk

Tell us in a few lines what you are working on, where it hurts and by when. We reply within one working day.

Get in touch

Or write to [email protected]