# nicojahn

> Independent applied-AI consulting for European enterprises

---

# About nicojahn

# About nicojahn

nicojahn is **Nico Jahn**, an independent applied-AI consultant based in **Berlin**. I
help European enterprises turn AI from a slide deck into production systems that ship,
scale, and stay compliant.

I'm a hands-on engineer and strategist, not a reseller. I write the code, own the
outcomes, and hand over systems your own people can run.

## What we do

- **[LLM & GenAI](./services/llm-genai.md)** — RAG, agents, fine-tuning, and evaluation pipelines.
- **[ML Engineering & MLOps](./services/mlops.md)** — model deployment, pipelines, and monitoring.
- **[AI Strategy](./services/ai-strategy.md)** — use-case discovery, roadmaps, and executive advisory.
- **[Data & Compliance](./services/data-compliance.md)** — data platforms and EU AI Act / GDPR readiness.

## Why European teams work with us

| | |
|---|---|
| **Production-first** | We measure success in deployed systems, not workshops. |
| **EU-native** | GDPR and the EU AI Act are designed in, not bolted on. |
| **Data sovereignty** | On-prem, EU-region cloud, or hybrid — your data stays where it must. |
| **Knowledge transfer** | We document, pair, and hand over. No lock-in. |

## How to start

Most engagements begin with a focused **AI Discovery** sprint. See our
[engagement model](./engagement-model.md) or [get in touch](./contact.md).

:::tip[For AI tools]
This site is AI-readable. Point your assistant at
[`/llms.txt`](https://nicojahn.pages.dev/llms.txt) or connect our
[MCP server](https://nicojahn.pages.dev/mcp).
:::

---

# LLM & Generative AI

# LLM & Generative AI

We design and ship generative-AI systems that hold up under real traffic, real data,
and real compliance review.

## Where we help

### Retrieval-augmented generation (RAG)
Grounded answers over your own knowledge base — with citations, access control, and
evaluation. We build the ingestion, chunking, retrieval, and re-ranking stack, then
prove quality with offline and online metrics.

### Agents & workflows
Tool-using agents that automate multi-step work: ticket triage, document processing,
internal copilots. We scope agency tightly, add guardrails, and keep a human in the
loop where it matters.

### Fine-tuning & adaptation
When prompting is not enough, we fine-tune or adapt open models on your domain data —
on infrastructure you control.

### Evaluation & guardrails
Every system ships with an eval harness: golden datasets, regression tests, and
production monitoring for hallucination, cost, and latency.

## Typical outcomes

- A support copilot that deflects 40%+ of tier-1 tickets with cited answers.
- A document-processing pipeline that cuts manual handling from hours to seconds.
- An internal RAG assistant deployed in your EU cloud region, GDPR-clean.

## How we build

```python
# A grounded answer is only as good as its evaluation.
# Every nicojahn RAG engagement ships with a regression eval suite.
from nicojahn.eval import GoldenSet, score

results = score(
    system="support-copilot",
    dataset=GoldenSet.load("tier1-tickets-v3"),
    metrics=["faithfulness", "answer_relevance", "citation_accuracy"],
)
assert results.faithfulness > 0.95  # gate the deploy on quality
```

We default to the most capable models for the task and keep the architecture
provider-flexible, so you are never locked to one vendor.

→ Next: [ML Engineering & MLOps](./mlops.md) · [Talk to us](../contact.md)

---

# ML Engineering & MLOps

# ML Engineering & MLOps

Models are easy to train and hard to keep alive. We build the engineering layer that
takes a notebook to a system your team can trust at 3 a.m.

## Where we help

### Deployment
Package models as versioned, observable services — batch, real-time, or streaming.
Kubernetes, serverless, or on-prem, matched to your stack and data-residency needs.

### Pipelines
Reproducible training and inference pipelines with data versioning, CI/CD, and
automated retraining triggered by drift or schedule.

### Monitoring
Track data drift, model quality, latency, and cost in production. Alerts that page a
human before customers notice.

### Platform
A paved road for your data scientists: feature store, model registry, and templates
so the next model ships in days, not quarters.

## What good looks like

```yaml
# Every model we ship is versioned, gated, and observable.
deploy:
  model: churn-predictor
  version: 2.4.1
  gates:
    - eval_auc: ">= 0.86"      # block regressions
    - latency_p99_ms: "< 150"
  monitor:
    drift: psi                  # population stability index
    alert_channel: slack#ml-ops
  region: eu-central-1          # data stays in the EU
```

## Typical outcomes

- Deploy time from weeks to a single merge.
- Drift caught and retrained automatically, not via customer complaints.
- A platform your own engineers extend without us.

→ Next: [AI Strategy](./ai-strategy.md) · [Talk to us](../contact.md)

---

# AI Strategy

# AI Strategy

Most AI budgets are spent on the wrong use cases. We help leadership find the few that
matter, sequence them, and build the capability to deliver.

## Where we help

### Use-case discovery
Structured workshops that map your processes to AI opportunities, scored by value,
feasibility, and risk. You leave with a ranked portfolio, not a wish list.

### Roadmap & business case
A phased plan with cost, expected return, and the data and platform prerequisites for
each step — defensible to your CFO and your board.

### Operating model
How AI teams sit alongside product and IT: build vs. buy, in-house vs. partner, and
the governance to keep it safe.

### Executive advisory
Ongoing sparring for CTOs and innovation leads navigating a fast-moving field —
including what the EU AI Act means for your roadmap.

## The discovery sprint

| Week | Focus | Output |
|---|---|---|
| 1 | Process & data mapping | Opportunity longlist |
| 2 | Scoring & prioritization | Ranked portfolio |
| 3 | Roadmap & business case | Phased plan + budget |

## Typical outcomes

- A board-ready AI roadmap with quantified ROI.
- Two to three de-risked first use cases ready to build.
- Clear build/buy and governance decisions.

→ Next: [Data & Compliance](./data-compliance.md) · [Engagement model](../engagement-model.md)

---

# Data & Compliance

# Data & Compliance

AI is only as trustworthy as the data and governance beneath it. We build the data
foundation and the compliance posture that European regulators — and your customers —
expect.

## Where we help

### Data platforms
Modern, EU-resident data infrastructure: ingestion, warehouse/lakehouse, governance,
and the quality controls AI depends on.

### EU AI Act readiness
Classify your systems by risk tier, identify obligations, and build the technical
documentation, logging, and human-oversight controls the Act requires.

### GDPR / DSGVO
Lawful-basis review, data-minimization, retention, and DPIAs for AI processing — so
personal data is handled correctly from ingestion to inference.

### Governance
Model cards, audit trails, and approval workflows that make AI decisions explainable
and defensible.

## EU AI Act, in brief

The EU AI Act tiers systems by risk. Most enterprise AI lands in **limited** or
**high** risk:

| Risk tier | Examples | Core obligations |
|---|---|---|
| Unacceptable | Social scoring | Prohibited |
| High | Hiring, credit scoring | Risk mgmt, docs, human oversight, logging |
| Limited | Chatbots | Transparency / disclosure |
| Minimal | Spam filters | None |

We map each of your systems to a tier and a concrete checklist — early, before it
becomes a launch blocker.

:::warning[Not legal advice]
We deliver the technical controls and documentation that support compliance. Pair us
with your legal counsel for binding interpretation.
:::

→ Back to [about nicojahn](../intro.md) · [Talk to us](../contact.md)

---

# Engagement Model

# Engagement Model

We keep engagements small, senior, and outcome-driven. Three phases, no open-ended
retainers unless you want one.

## 1. Discover — 2 to 3 weeks

A fixed-scope sprint to find and de-risk the right use cases. You get a ranked
portfolio, a roadmap, and a business case. Fixed price, no commitment beyond it.

## 2. Build — 6 to 12 weeks

A senior pod (typically 2–4 engineers plus a lead) ships a production system in two-week
increments. You see working software every sprint, deployed to your environment.

## 3. Operate & hand over

We instrument, document, and pair with your team until they can run it without us.
Optional ongoing support if you want a safety net.

## Principles

- **Senior-only delivery.** The people who scope are the people who build.
- **Your infrastructure.** We deploy into your cloud or on-prem. Your data stays yours.
- **No lock-in.** Open standards, full documentation, knowledge transfer by default.
- **EU-native.** GDPR and EU AI Act considerations are part of every phase.

## What we need from you

- A business owner who can make decisions.
- Access to relevant data and systems (under your governance).
- One or two of your engineers to pair with — so the capability stays in-house.

Ready? [Get in touch](./contact.md).

---

# Contact

# Contact

Let's talk about where AI can move the needle for your business.

## Get in touch

- **Email:** [nico.k.jahn@gmail.com](mailto:nico.k.jahn@gmail.com)
- **Book a call:** a free 30-minute AI discovery call
- **Office:** Nico Jahn · Berlin

We typically reply within one business day.

## For AI tools

This documentation is machine-readable. Connect your assistant directly:

- **MCP server:** `https://nicojahn.pages.dev/mcp` — works with Claude, Cursor, Windsurf, and any MCP-compatible client.
- **llms.txt:** [`https://nicojahn.pages.dev/llms.txt`](https://nicojahn.pages.dev/llms.txt)
- **Full context:** [`https://nicojahn.pages.dev/llms-full.txt`](https://nicojahn.pages.dev/llms-full.txt)

### Claude Code

```json
{
  "mcpServers": {
    "nicojahn": { "type": "http", "url": "https://nicojahn.pages.dev/mcp" }
  }
}
```