AI systems consultancy

Ship AI that actually holds up in production.

Your team is staring at a pile of LLM prototypes, a hype-allergic CTO, and a launch date. You want answers that survive real traffic — not another slide deck. Good. That's what gets built here. A small senior team plugs into yours for a few focused weeks, removes the vague bits, and leaves you with a system your engineers can run on Monday morning.

If this sounds familiar

You don't need another deck. You need someone who ships.

You've run the workshops. You've read the threads. You've watched three vendors demo the same chatbot with different names on the tab. Meanwhile the actual product still hallucinates, the retrieval is flaky, costs are creeping, and nobody on your team has the bandwidth to fix it properly without falling behind on the roadmap.

That's the spot we come in. Not to tell you AI is the future — you know. To sit at your workbench for a few weeks, get hands dirty, and hand you something testable, observable, and honest.

What you actually get

Six outcomes we solve for — in your codebase, not a sandbox.

01

A working LLM feature, wired into your product

Not a notebook. A shipped path: retrieval, evals, guardrails, rollout plan. Your engineers own it by the end.

02

An agent that survives its second week

Tool-use, memory, traces, and the boring parts — budgets, fallbacks, replay. Built to debug, not to demo.

03

Eval harness that actually catches regressions

Golden sets, scorer rubrics, CI hooks. Stops the "it worked on my prompt" cycle before launch.

04

Cost and latency under control

Model routing, caching, streaming, batching. We'll show the math before touching production.

05

Your team, levelled up while we work

Pairing over presentations. When we leave, your engineers have the muscle — and the notes — to keep shipping.

06

A straight answer about what's real

If a problem doesn't need an LLM, we say so. If it does, we tell you where the dragons live. No hand-waving.

Ways to work

Five shapes. Pick the one that matches your mess.

01 — Discovery Call

60 minutes · free

We hear your context, tell you straight whether we can help, and share a written sketch of what a pilot would look like.

02 — Pilot Week

5 days · fixed scope

One week, one question: is this AI idea worth building? You leave with a prototype, a cost model, and a "keep going / stop" recommendation.

03 — Build Sprint

4–8 weeks

A focused team sits next to yours and ships a feature end-to-end. Retrieval, agents, evals, observability — whatever the thing needs.

04 — Embedded Advisor

1–2 days / week · ongoing

Your team keeps the keyboard. We review PRs, unblock tough calls, and keep the architecture from drifting into fashion.

05 — Hands-on Engineering Workshop

2 days · on-site or remote

Your engineers build a real thing in our codebase: an agent, a RAG pipeline, or an eval harness. No slideware. Working code at the end.

Proof over promises

What the workshop has turned out.

Open source · 15+ years

Libraries engineers actually use

Vide (incremental UI), FsHttp (HTTP DSL), Trulla (templating), LocSta (state machines), FluX (data flow). Not weekend hacks — published, maintained, tested in production by teams we've never met.

  • github.com/SchlenkR — active, public, reviewable
  • .NET / F# / C# — deep, not decorative
Product · Cumin & Potato GmbH

PXL Clock — hardware + cloud + SDK

A 24×24 RGB display programmable in C#, with a browser simulator, OTA firmware, a mobile app, and a community of makers. Two people built it. That's the operating model we bring to your team: small, senior, accountable.

  • End-to-end shipping across hardware, cloud, mobile, DX
  • Documented, versioned, and maintained — not a demo
Honest answers

The questions everyone asks on the second call.

We're not "AI-ready." Are we wasting your time?
If you have a product, a codebase, and people who care about it — you're ready enough. Half the pilot is cutting the idea down to something worth shipping. That works better when the ground isn't already paved.
Why not hire a full-time ML team?
Eventually, probably. But hiring takes six months and you need an answer this quarter. A pilot tells you what shape of team you actually need — so the eventual hire lands on a clear brief, not vibes.
We've been burned by consultants before.
Fair. Two differences: we commit working code, not decks; and we default to leaving. If after week two you don't see it on the board, you close the engagement. No retainer trap.
Our stack is .NET / F# / C#. Is that a problem?
The opposite. That's home turf — 15+ years. We can speak to your engineers in their language, and we won't try to Pythonize your codebase because it's trendier.
Can we self-host everything? No cloud vendors?
Yes. We have production experience running local models and self-hosted inference — not just lab toys. If data gravity or regulation says "no cloud," we can plan for that from day one.
How small is "small team"?
Two senior engineers, on average. Sometimes three. Never a ladder of juniors billed at a senior rate. You meet the people doing the work in the first call.
The person at the workbench

Ronald Schlenker — and a very small team.

Fifteen-plus years building software that ends up in production: distributed systems, SDKs, compilers, and — for the last few — LLM pipelines, agents, and retrieval-heavy features. Mostly .NET / F# / C#. Public body of work on GitHub. A habit of finishing.

Alongside consulting, Ronald and one other engineer run Cumin & Potato GmbH — the company behind PXL Clock. That's not a side note; it's the operating model. Small. Hands-on. Allergic to decks that outweigh the code.

Based
Germany · remote-first
Stack
.NET · F# · C#
Focus
LLM systems · agents
Start small. Prove it. Decide.

Book a pilot week and know by Friday.

One week. One question answered. One working prototype. If it's not worth building, we say so — and you'll have saved a quarter.