Service / 06

Artificial intelligence

We build chatbots, integrate model APIs and run models on your own infrastructure — with retrieval, evaluation, access controls and training for the people using them.

Start a project

QUERY

RETRIEVE

CONTEXT

MODELgpt · finetuned

ANSWER

grounding window 12k tokens · rerank on · guardrails on

embeddingsvector storererankereval harness

FIELD NOTE / 01

AI needs an engineering system around it.

30% of developers have little or no trust in AI-generated code — DORA, 2026. Evaluation, human review and operational ownership belong in the design from day one.

Read DORA’s research

The work

Designed for
the conditions.

01Chatbot integration
02AI training
03Local AI (on-premise)
04LLM integration
05Retrieval augmented generation
06Embeddings and vector search
07AI APIs
08Evaluation and automation

Process

Clear steps,
visible progress.

Frame the use case
Decide where a model genuinely helps the workflow.
Ground in data
Connect retrieval, permissions and business context.
Evaluate continuously
Measure answer quality before and after release.
Operate with controls
Guardrails, human review and cost visibility.

Operating targets

Evaluated answers

Answer quality

Per-request costs

Cost visibility

Access controlled

Security

Live example

Edit the code. Run it. Break it.

A real, runnable slice of this work — your edits run in a sandboxed preview.

playground/artificial-intelligence.html

<h2>deploy · main → production</h2>
<p class="hint">The pipeline streams its own logs as it runs — this whole window is one deploy, from commit to edge.</p>
<div class="grid">
  <div class="stages">
    <div class="stage" id="st0"><i></i><span>BUILD</span><small></small></div>
    <div class="stage" id="st1"><i></i><span>TEST</span><small></small></div>
    <div class="stage" id="st2"><i></i><span>SHIP</span><small></small></div>
  </div>
  <div class="term"><pre id="log"><span class="cursor"></span></pre></div>
</div>
<div class="tiles">
  <div><b>12</b><small>edge regions</small></div>
  <div><b>1.2k</b><small>req / s</small></div>
  <div><b>84 ms</b><small>p95 latency</small></div>
  <div><b id="runs">1</b><small>deploys today</small></div>
</div>
<p id="status" class="idle">waiting for push…</p>
<button id="again" class="ghost">Deploy again</button>

⎇ main0 errors · 0 warningsLn 1, Col 1 · Spaces: 2 · UTF-8 · HTML · ○ idle

sandbox.nordiso.local/artificial-intelligence

sandbox idle — boots when this section scrolls into view

Technology examples

Model APIsVector searchEmbeddingsData pipelines

TypeScript

Examples, not a fixed stack — chosen per project, with the team that will own it.

TypeScript

Have something important to build?

Let’s make a good start.

Start a project

<h2>deploy · main → production</h2> <p class="hint">The pipeline streams its own logs as it runs — this whole window is one deploy, from commit to edge.</p> <div class="grid"> <div class="stages"> <div class="stage" id="st0"><i></i><span>BUILD</span><small></small></div> <div class="stage" id="st1"><i></i><span>TEST</span><small></small></div> <div class="stage" id="st2"><i></i><span>SHIP</span><small></small></div> </div> <div class="term"><pre id="log"><span class="cursor"></span></pre></div> </div> <div class="tiles"> <div><b>12</b><small>edge regions</small></div> <div><b>1.2k</b><small>req / s</small></div> <div><b>84 ms</b><small>p95 latency</small></div> <div><b id="runs">1</b><small>deploys today</small></div> </div> <p id="status" class="idle">waiting for push…</p> <button id="again" class="ghost">Deploy again</button>

Artificial intelligence

AI needs an engineering system around it.

Designed forthe conditions.

Clear steps,visible progress.

Frame the use case

Ground in data

Evaluate continuously

Operate with controls

Edit the code. Run it. Break it.

Let’s make a good start.

Artificial intelligence

AI needs an engineering system around it.

Designed forthe conditions.

Clear steps,visible progress.

Frame the use case

Ground in data

Evaluate continuously

Operate with controls

Edit the code. Run it. Break it.

Let’s make a good start.

Designed for
the conditions.

Clear steps,
visible progress.

Designed for
the conditions.

Clear steps,
visible progress.