Skip to content

Software development · Accelerators · Europe

Software that startshalfway built.

We build custom software

OpsVector is a European software development company. Every project starts from an accelerator — working code and written-down decisions for the parts everyone rebuilds — so your budget goes on the part only you have.

Small senior team · Working across the EU · Taking on new work
  • Retrieval-augmented generation
  • Model Context Protocol
  • Document extraction
  • LLM evaluation
  • Agentic workflows
  • AWS
  • OpenTofu
  • Terragrunt
  • TypeScript
  • Infrastructure as code
  • Observability
  • GDPR
  • EU AI Act

In one paragraph

OpsVector is a European software development company. We build custom software, AI systems and cloud infrastructure from reusable accelerators — retrieval (RAG), Model Context Protocol (MCP) servers, document extraction, LLM evaluation and AWS landing zones — adapted to each client. The work is designed around GDPR, the EU AI Act and your own access policy.

Who we are

We are a small senior team that builds software meant to work on its first serious day, not just on demo day. Instead of opening every project with an empty repository, we start from foundations we have already thought hard about — and spend your time on what makes your problem yours.

Custom software development, AI engineering and DevOps for small and mid-sized European companies.

How we work

01Services

Five things we build, end to end.

Custom software development, AI and LLM engineering, cloud and DevOps, data and document pipelines, and modernisation of existing systems.

  1. S/001

    Custom software

    Web applications, internal tools and APIs shaped around how your team actually works. Typed, tested, deployed by pipeline, and documented well enough that your own engineers can take it over.

    • Web applications
    • Internal tools
    • APIs
  2. S/002

    AI & LLM engineering

    Retrieval over your own documents, agents with narrow and audited tools, and the evaluation that tells you whether a change made things better or merely different.

    • RAG
    • MCP servers
    • Agents
    • Evaluation
  3. S/003

    Cloud & DevOps

    AWS infrastructure as code with OpenTofu and Terragrunt, delivery pipelines, observability, and a rollback you have actually rehearsed.

    • AWS
    • OpenTofu
    • CI/CD
    • Observability
  4. S/004

    Data & documents

    Classification, extraction and validation for invoices, contracts and identity documents, with a person in the loop wherever a mistake would cost more than the review.

    • OCR
    • Extraction
    • Validation
  5. S/005

    Modernisation

    Moving an ageing system onto ground you can change safely: pin down what it does today, then replace it a piece at a time while it keeps serving users.

    • Legacy systems
    • Characterisation tests
    • Incremental delivery

02Accelerators

Start from a foundation, not a blank page.

Reusable accelerators for RAG, MCP servers, document extraction, LLM evaluation, AWS landing zones and legacy modernisation.

All accelerators

An accelerator is a foundation we bring into your project: working code, infrastructure and the design decisions behind them, already written down. We adapt it to your systems and you own the result. None of them is a product you subscribe to.

A/01

Governed RAG Foundation

Answers from your documents — only from the ones the person asking may read.

Retrieval-augmented generation over company data, with access control carried from the source system and a citation on every answer.

  • RAG
  • Vector search
  • Access control
  • Citations
Explore
A/02

MCP Server Kit

Give agents narrow, typed tools — never a shell.

A starting point for Model Context Protocol servers with a small, typed, read-first tool surface, a log of every call, and an approval step for anything that changes state.

  • MCP
  • Agents
  • Audit trail
  • Least privilege
Explore
A/03

Document Extraction Pipeline

From a scanned page to validated fields, with a person where it matters.

Classification, extraction and validation for business documents — invoices, contracts, identity documents — with a review queue for results the system is unsure of.

  • OCR
  • LLM extraction
  • Schema validation
  • Human review
Explore
A/04

LLM Evaluation Harness

“Is it better?” — with an answer that is a number.

Regression testing for prompts, models and retrieval: a versioned set of cases drawn from your own history, scored on every change.

  • Evaluation
  • CI
  • Regression testing
  • Model comparison
Explore
A/05

AWS Landing Zone

Infrastructure as code from the first commit.

An OpenTofu and Terragrunt foundation for AWS: an explicit dependency graph, a pinned toolchain in a container, and CloudFront hosting behind a strict Content-Security-Policy. This website is deployed on it.

  • AWS
  • OpenTofu
  • Terragrunt
  • CloudFront
Explore
A/06

Modernisation Starter

Change an old system without switching it off.

A toolkit for moving an ageing application forward safely: an inventory of what it really does, tests that pin its behaviour, and scaffolding to replace it piece by piece.

  • Legacy systems
  • Characterisation tests
  • Strangler fig
  • Incremental delivery
Explore

03Why accelerators

The first weeks go to your problem.

What changes when a software project starts from an accelerator instead of an empty repository.

  • 01

    Working code on day one

    Authentication, pipelines, logging and infrastructure are already there to adapt, so early effort lands on the part that is specific to you.

  • 02

    Decisions you can read first

    Each accelerator comes with the reasoning behind it, including the options we rejected. You can review the design before committing to it.

  • 03

    You own the result

    The code is delivered into repositories you control. No licence fee, and no platform you have to keep paying for to keep it running.

  • 04

    Fits the stack you have

    Accelerators are starting points, not frameworks. We adapt them to your systems, conventions and constraints rather than asking you to adopt ours.

  • 05

    European defaults

    Data residency, access control and audit records are part of the foundation, so GDPR and EU AI Act questions have answers from the start.

  • 06

    Familiar to whoever comes next

    The same patterns across projects mean the documentation, runbooks and tests look familiar to the next engineer who has to maintain them.

04Two ways to start

Same project. Two very different first months.

Building custom software from an empty repository versus from a reviewed accelerator.

Take a company that wants an assistant to answer staff questions from internal documents. Here is how the two starts differ.

FROM AN EMPTY REPOSITORY

The team builds the plumbing first: ingestion, a vector store, sign-in, deployment. A demo answers questions within days.

  • index every document into one store
  • add permissions later
  • judge quality by trying a few questions
  • deploy by hand to see it working

Then the hard questions arrive. Who may see which answer? Did last week’s prompt change break anything? How exactly is it deployed? Each answer means reopening a decision made in a hurry.

FROM AN ACCELERATOR

The plumbing exists and its decisions are already written down. The first weeks go into your sources, your permissions and your questions.

  • access lists carried from the source system
  • evaluation set built from your questions
  • every answer cites its source
  • infrastructure as code from the first commit

The hard questions still arrive. They simply have answers already, and the evidence to go with them.

An accelerator does not make a project small. It moves the effort from re-solving common problems to the part only you have.

05Who we work with

For teams at different starting points.

Software development for startups, scale-ups, regulated companies and in-house engineering teams in Europe.

06How we work

Four steps. You can stop after any of them.

How a software development engagement runs, from discovery through to handover to your own team.

Nobody should have to commit to a six-month build before finding out whether the idea holds. Each stage produces something you keep and could hand to somebody else.

  1. STEP 01

    Discovery

    A short engagement, usually days rather than weeks. We talk to the people doing the work, not only the people describing it, and come back with a written problem statement and an honest verdict — sometimes that something simpler would do.

  2. STEP 02

    Fit the accelerator

    We choose the foundations that fit, and write down what has to change for your systems, where your data travels and who may see it. If none fits, we say so and design from scratch. The document is yours either way.

  3. STEP 03

    Build

    The smallest version that proves the idea, on real data, under real constraints, running where it will eventually live. Measured against cases from your own history, so “is it good enough?” has a numeric answer.

  4. STEP 04

    Hand over

    Infrastructure as code, pipelines, monitoring and runbooks, with your engineers walked through all of it until they can change it without us. A supplier you cannot leave is a risk on your register, not a service.

07Values

How we build, whatever we build.

The engineering principles behind every OpsVector project.

  1. 01

    Narrow tools, not shells

    A tool that can do anything can do anything wrong. We list the actions a system genuinely needs, type them, bound them, and expose nothing else.

  2. 02

    Code owns what must be exact

    Language models paraphrase, and paraphrasing a number is a liability. Code computes every figure; the model writes only the prose around it.

  3. 03

    Every input is untrusted

    A document can carry an instruction and a folder can hold a secret. Retrieval and file access are attack surfaces, and we build them like ones.

  4. 04

    If you can’t reconstruct it, you can’t defend it

    Which inputs, which sources, which version, which person approved it. We record that from the first day, because it cannot be added later.

08In the open

No testimonials yet. Read our reasoning instead.

OpsVector is a young company, and we will not decorate this page with logos we have not earned. What we can show is how we think: the design path of real systems, including the turns that were wrong and why we took them.

Judge the reasoning. If it holds up, the work will too.

09Questions

What people ask before they write.

Common questions about accelerators, software development engagements, code ownership and data residency.

What exactly is an accelerator?

A foundation we bring into your project: working code, infrastructure definitions and written design decisions for a problem that comes up again and again, such as retrieval with access control or an AWS landing zone. We adapt it to your systems, and it becomes part of your codebase rather than a product you subscribe to.

Do we own the code?

Yes. What we build for you is delivered into your repositories and is yours to change, extend or hand to another supplier. There is no licence fee to keep it running.

What if none of the accelerators fits?

Then we build from scratch and say so up front. Accelerators exist to save time where a problem is common; forcing one onto a problem it does not fit would cost more than it saves.

Where does our data actually go?

Wherever you decide, and you will know before anything moves. Some work can run entirely on infrastructure you control, with models that never leave it; some is better served by a hosted provider. Which system sees which data, in which jurisdiction and under whose terms is a design decision we put in writing, not a detail discovered afterwards.

You’re new. Why would we trust you with this?

On the evidence, not on our say-so. OpsVector is a young company with a small senior team, and we will not pretend otherwise. Start with the smallest stage, judge the written output, and continue only if it is good. Everything we produce is documented so it stays useful even if you never work with us again.

How does the commercial side work?

Each stage is scoped and agreed before it begins, so there is no open-ended commitment and no surprise at the end of a quarter. What a stage costs depends on what it contains, which is why the conversation starts with the problem rather than a price list.

Start a project

Tell us what you want to build.

Describe it in plain words: what it should do, who will use it, what you have already tried. If an accelerator fits, we will say which one. If nothing we do fits, we will say that too.

No form and no automated follow-up sequence. An email reaches an engineer, and the first reply is either a real question or an honest “not us.” We work in English, remotely, across European time zones.

contact@opsvector.ai

We reply to every message that isn’t automated.