Skip to contact
NANDO THEESSEN
DE EN
Book a call

Nando Theessen · Infrastructure engineer for Kubernetes, cloud cost and data platforms

I build infrastructure that holds up in production.

I have worked for many years on systems that have to stay up: in German healthcare, on the electricity grid and most recently for a defence technology startup. My focus is Kubernetes, cloud cost, data platforms and private AI, and faults whose cause is not obvious.

You bring me in when infrastructure gets difficult, a migration stalls or a problem has to be solved quickly. Architecture and technical coordination run through me. When a project grows, I bring in engineers I have worked with. I stay accountable.

nando@devwire.technology

Nando Theessen

01 / RESULTS

The difference is measurable.

99.8%lower translation API costs3 billionmedical documents migrated in under 2 days2 weeksto diagnosis and a replacement plan~€10,000less infrastructure spend per month

02 / SELECTED PROJECTS

Engineering that changes the outcome.

Fewer outages. Lower running costs. Delivery that keeps moving. Six projects and the work behind the results.

PRIVATE AI · ONGOING

Confidential defence technology company

From rationed translation to private AI at scale.

Translation costs were limiting how much source material the team could process. DevWire built a private inference gateway on dedicated infrastructure, then expanded it to speech, vision, sentiment analysis and internal LLMs.

The team can now process more material, with sensitive AI workloads running on its own infrastructure.

99.8%lower Google Translate API costs
Former monthly peak€5,600
Monthly API bill after€13

500,000+inference requests per day

€2,400monthly hosting for the entire expanded private AI platform

The 99.8% reduction applies to the Google Translate bill. Platform hosting is a separate cost.

01 / HEALTHCARE

Barmer & HEK · via T-Systems

3 billionmedical documents migrated in under two days

Ready for the ePA deadline.

An 18-month engagement, with one DevWire architect leading a team of six. The platform delivered 30,000 reads per second and launched for the 15 January 2025 deadline.

Inside the project

The challenge. Sensitive medical records, a fixed launch date and a growing shared application platform made this more than a database migration.

The work. DevWire selected and tuned MongoDB on Kubernetes, led the six-person team and delivered the initial migration at roughly 21,000 documents per second across more than ten shards.

The result. The insurers launched on time on a shared, monitored platform. T-Systems engaged DevWire for further projects.

02 / RELIABILITY

Rackspace · end customer beck-online

2 weeksto root cause and a replacement plan

Months of outages. A cause finally found.

A stalled container engine left monitoring reporting stale, healthy status. DevWire identified the failure, then replaced the engine live during the two-month engagement.

Inside the project

The challenge. Recurring Windows node failures disrupted beck-online, the C.H. Beck legal research service, while monitoring still showed healthy systems. Senior engineering teams had investigated for months.

The work. One DevWire staff engineer traced the failure to the officially supported container engine. The diagnosis and replacement plan were ready in two weeks; the live replacement followed during the two-month engagement.

The result. The outages stopped after the engine replacement. Rackspace retained the customer, maintained its service commitments and freed senior engineers from the escalation.

03 / PLATFORM OPERATIONS

Confidential defence technology company

~€10,000saved on infrastructure each month

Rebuild the platform. Keep shipping.

DevWire took over an inherited platform, rebuilt roughly half of it while releases continued, and cut cloud costs approximately in half. Ongoing work by an architect and a senior engineer.

Inside the project

The challenge. The previous platform engineer had left. Releases were unpredictable, knowledge was incomplete, internal services were exposed to the internet and the data layer was costly.

The work. DevWire resolved recurring failures, introduced controlled access, secrets and identity management, monitoring and tested recovery. A self-hosted Elasticsearch and ClickHouse data layer supports around 50 TB.

The result. Right-sizing and committed-use discounts reduced recurring infrastructure costs by around €10,000 a month. DevWire continues to operate the platform and take responsibility for its availability.

04 / DEVELOPER PLATFORMS

50Hertz

70specialists on one shared developer platform

Get engineering moving. Then bring it in-house.

One DevWire architect led the platform team over a year. Development started in the public cloud while in-house infrastructure was prepared, then migrated successfully.

Inside the project

The challenge. The grid digitalisation programme brought together specialists from Accenture, NTT and smaller consultancies. At the first hiring round, shared development tools and environments were still missing.

The work. DevWire established source control, task tracking, documentation and secure development environments, leading engineers from Evoila and independent specialists.

The result. Seventy specialists gained a common platform and working onboarding. The platform supports delivery into an air-gapped environment.

05 / SOFTWARE VALIDATION

Audi

Less manual prepmore repeatable vehicle software validation

Configure. Test. Get feedback.

Over a year, one DevWire architect connected software pipelines to physical test hardware and emulators. Vehicle configurations and test suites became selectable parts of a repeatable workflow.

Inside the project

The challenge. More hardware and vehicle models meant hours of manual preparation: find the device, configure it, install software, run tests and publish results.

The work. The workflow automates flashing and preparation, installation, functional, smoke, unit and end-to-end tests, and reporting.

The result. Used daily, the workflow gives engineers earlier feedback with less preparation and reporting work. New variants can be added through configuration.

The two confidential projects are for the same defence technology client.

FROM THE HEALTHCARE PROJECT

“Their professionalism and collaborative spirit made them an indispensable part of our success, even when faced with the most demanding challenges.”
Norbert V.Service Delivery Manager, T-Systems

Have a similar challenge?

Let’s talk about it

03 / WHAT I DO

What you can bring me in for

01

Kubernetes and developer platforms

When your clusters have grown over years, a platform needs rebuilding, or a migration has stalled.

The work. I design the target architecture, build or migrate clusters and set up GitOps, delivery, namespaces, access rights, autoscaling and observability.

Aim. Teams deploy on their own without every release becoming an ops ticket.

02

Cloud cost

When the cloud bill grows faster than usage and it is unclear which workloads are driving the cost.

The work. I break the cost down by workload, re-size resources, remove what is unused and review commitments, discounts, storage and traffic. Then I implement the changes.

Aim. Running cost falls at the same or better performance. For one previous client that came to around €10,000 per month.

03

Data platforms and databases

When queries are slowing down, a migration is due, or a database becomes unstable under load.

The work. I work on PostgreSQL, MongoDB, Elasticsearch and ClickHouse: sharding, replication, performance analysis, backup and restore, and migrations with a controlled cutover.

Aim. The data layer stays stable and predictable under real load.

04

Private AI infrastructure

When inference needs to move onto your own infrastructure because data cannot leave the company, or API costs are rising too steeply.

The work. I build the infrastructure for it: GPU nodes, scheduling, model serving, scaling and operational procedures for language, speech, vision, embeddings and LLM inference.

Aim. The models run in your environment. Data paths and running costs stay transparent. Running it yourself is not free, but it is controllable.

05

Troubleshooting and infrastructure security

When the same fault keeps coming back, or security hardening stays stuck in the backlog.

The work. I find the cause and implement the corrections. I harden clusters, secure network paths, clean up access rights and secrets, and create dependable routes for patches and updates.

Aim. The problem is fixed at the cause, and the protection fits your environment. At Rackspace the diagnosis and replacement plan were ready in two weeks. Implementation followed during the two-month engagement.

06

Ongoing operations and development

When the platform needs to be operated and developed further after the project.

The work. We work with clear responsibilities, agreed availability and a model that fits your team.

Aim. The platform stays maintained and keeps developing, even when internal capacity is short.

Book a free 20-minute intro call

Your topic does not fit cleanly into one of these categories? Write anyway: nando@devwire.technology

Nando Theessen

04 / WHO I AM

Nando Theessen

I have worked for many years on systems that have to stay up: in German healthcare, on the electricity grid at 50Hertz and most recently for a confidential startup in defence technology. In environments like these, technical decisions have to be traceable.

You work directly with me. Architecture, technical decisions and coordination run through me. When scope or pace calls for it, I bring in engineers I have worked with. I stay accountable. You are not building a new team and you are not hiring a large consultancy.

You get a clear assessment, even when it argues against the work. Decisions have to be technically justifiable. I pick technology for what it does, not because it is in fashion, and I hand over so your team can carry on.

FOCUS

Kubernetes, cloud cost, data platforms, private AI infrastructure, infrastructure security

SECTORS

Healthcare, energy and grid operation, defence technology, automotive, product companies and startups

HOW I WORK

Remote, in German and English, inside your existing ways of working

05 / HOW IT WORKS

How we work together

  1. STEP 01

    20 minutes, free

    We talk about your environment, the goal, rough scope and timing. Afterwards you know whether I can help and what the sensible next step is.

  2. STEP 02

    Agree the goal and scope

    We fix which outcome should be reached first, at what budget and within what boundaries. Investigation and delivery are paid work.

  3. STEP 03

    Delivery

    I work in your environment and with your team on an agreed rhythm. Progress and decisions stay visible.

  4. STEP 04

    Handover or continued work

    Either I hand over to your team with documentation, or we continue with clear responsibilities. Both are agreed deliberately.

Scope, working rhythm and price are agreed per project to fit your situation.

06 / COMMON QUESTIONS

Before you write

If your question is not here, ask it on the call or by email.

Do you work remotely?

Yes. By default I work remotely from Germany. For workshops, sensitive environments or restricted access I come on site when needed.

We have our own ops team. Does that work?

That is the normal case. I do not replace your team, I take on a clearly bounded part: a platform, a migration or a recurring problem. We clarify responsibilities and decision paths at the start.

We are not sure where the problem actually is.

Just describe the symptoms: deployments take too long, the bill is rising, the database becomes unstable, or ownership is unclear. On the call we narrow down what is worth pursuing. A dependable investigation is then a separate, paid step.

Project work or ongoing support?

Both. Many engagements start as a clearly bounded project and then continue as ongoing support. Responsibilities, availability and scope are then agreed explicitly.

Our environment is sensitive.

Yes. I have worked in regulated and confidential environments, including German healthcare and defence technology. NDAs, restricted access, review processes and working solely inside your infrastructure are familiar to me. I discuss specific clients only within what has been agreed.

What is included in the free call?

Twenty minutes on your environment, the goal, rough scope and a sensible next step. Afterwards you know whether I am a fit for the topic. A technical investigation of your environment is then the first paid step and is agreed beforehand.

07 / CONTACT

Tell me what is not working.

You reach me directly, by appointment or email. No sales team in between.

RECOMMENDED

Free 20-minute intro call

Your environment, your goal, rough scope and timing, plus a sensible next step. No deck, no sales pitch.

Pick a time in the calendar

Opens the calendar in a new tab. If booking does not work for you, just send an email instead.

Prefer to write?

A few sentences about your environment and the problem are enough for a first read on it.

nando@devwire.technology

English and German both work. If German suits you better, there is a German version of this page.