Guztia AI Infra Ops

Guztia /How it works

The approach, at a useful altitude

How It Works

What we actually build, described by function rather than by brand. Also: why we do not publish the product choices, where open-weight models genuinely still lose, and what you get to see before committing to anything.

Book an assessment What it costs

01 The shape of it

Guztia reference architecture for enterprise AI Engineer tooling, product runtimes and CI pipelines all send requests to a gateway you control. The gateway routes them to open-weight models running on capacity in your own name, keeping a capped route to a premium vendor for the work that needs it. Every request is recorded underneath for cost attribution and audit. YOUR PEOPLE & PIPELINES Assistants & editors UNCHANGED CONFIGURATION Product runtimes YOUR APPLICATIONS CI & batch jobs AUTOMATED WORKLOADS CONTROL POINT — YOU OWN IT AI gateway ROUTING + FAILOVER SPEND LIMITS PER TEAM RATE LIMITS IDENTITY + ACCESS CACHING DATA-CLASS POLICY CHANGEABLE WITHOUT TOUCHING A LAPTOP Primary model · your capacity FIXED PRICE · REGION OF YOUR CHOICE Secondary model · your capacity FOR THE HARDER WORK Premium vendor · fallback CAPPED, MEASURED, OPTIONAL Cost & usage record — self-hosted EVERY REQUEST LOGGED · COST PER TEAM, PROJECT AND PERSON · LATENCY · QUALITY SCORES · RETAINED AUDIT TRAIL RUNS ON INFRASTRUCTURE IN YOUR NAME — AWS · ALIBABA CLOUD · GCP · AZURE · TENCENT · BARE METAL OPERATED BY GUZTIA UNDER A MONTHLY AGREEMENT — KEPT CURRENT · CAPACITY REBALANCED · INCIDENTS · MONTHLY COST REPORT
Fig. 01 — The gateway is the seam: left of it nothing changes, right of it everything can.

02 Four layers, all open source, all yours

Named by what they do, not by who makes them.

Every layer below is established open-source software. None of it is ours, none of it is licensed to you by us, and all of it runs under accounts in your name.

01 The gateway A single endpoint every team talks to instead of talking to a supplier directly. Routing, failover, spending limits per team, and the point at which switching supplier stops being an engineering programme. Open source
02 The record Every request logged with who asked, what answered, how long it took and what it cost. This is what turns one unexplained invoice into a breakdown you can charge back and defend. Open source
03 The serving layer What turns capacity you have bought into work delivered. How well this is configured decides your cost per unit of work, and it is where most self-assembled attempts quietly lose their savings. Open source
04 The models Open-weight models, run on infrastructure in your name. Good enough for the large majority of ordinary business work, and improving faster than the premium alternatives are getting cheaper. Open source

03 Why this page names no products

The architecture is not the secret. The judgement is.

Anyone can draw the diagram above; it is not novel and we have never claimed it was. What is worth paying for is knowing which of your workloads can move without anybody noticing and which cannot, how to size committed capacity against demand that has not settled yet, where these arrangements fail at two in the morning, and what the right mix is for a company shaped like yours.

Publishing that as a tutorial would be an odd way to run a firm. It is also how organisations end up with a half-built version assembled from a blog post, which we then get called in to take apart — usually after it has quietly lost the savings it was supposed to produce.

So the detail is not withheld from you. It is withheld from the open web.

WHAT YOU SEE BEFORE SIGNING

Everything that bears on the decision

  • The complete proposed design for your organisation, under NDA, before any commitment
  • A working demonstration — the actual reporting and controls, not slides
  • Your own numbers modelled, with every assumption written down and open to challenge
  • Your engineering lead in the room for as long as they want, asking whatever they like
  • A written statement of what you would own and what leaving would involve

No client has been asked to buy this on trust. The technical interrogation is welcome — it is usually where the conversation becomes serious.

Book an assessment

04 Why open source, specifically

Because the point is that you can fire us.

A managed service only its vendor can operate is not a solution to supplier lock-in. It is a smaller supplier with the same problem.

Everything we deploy is permissively licensed, runs in your accounts, and is documented publicly by people who are not us. That is not altruism — it is the only version of this business we think survives contact with a competent CTO, and it is why we charge for operating the arrangement rather than for access to it.

Not in the arrangement
Any Guztia-built software, any proprietary control layer, any margin on your usage, any hosted service of ours that your traffic must pass through.
You hold
The cloud accounts, the infrastructure code, the model weights, the usage records, the documentation.
We hold
A monthly invoice and the pager.
Leaving
Thirty days' notice. Nothing to extract, nothing to rebuild, no data to get back.

05 Where open weights still lose

We are not going to tell you the gap has closed.

It has narrowed a great deal. For the bulk of enterprise work — classification, extraction, summarisation, internal question answering, routine code — a good open-weight model is not distinguishable from a premium API in outcome, while costing a fraction. That is most of your volume, and it is the whole basis of the argument.

It is not all of it. There remain tasks where the premium models are simply better, and pretending otherwise would cost you more than you saved. That is exactly why every arrangement we build keeps a capped route to a premium vendor rather than burning the bridge.

Which of your workloads fall on which side is an empirical question about your work, not a matter of opinion — and answering it is the first thing an assessment does.

Long-horizon agents
Tasks running many steps with tool use and self-correction. Premium models hold together noticeably longer before drifting.
Genuinely novel reasoning
Unfamiliar problems rather than variations on well-covered ground. The gap is real and shows up clearly in testing.
Very long context
Open models advertise large windows; quality at the far end of them is a separate question. Test at the length you will actually use.
Low-resource languages
Coverage varies sharply. If you operate in one, it has to be tested specifically — a headline multilingual claim will not answer you.
Harder multimodal work
Open vision-language models are improving quickly but premium APIs remain ahead on difficult document and image reasoning.

THE RECOMMENDATION

Route by difficulty, not by loyalty

Bulk volume goes to capacity you control. Harder work goes to a larger model you also control. What neither can do well goes to a premium vendor, under a hard ceiling — and every one of those requests is recorded, so you can see precisely what the escape hatch costs and whether it is shrinking. In the arrangements we run, it shrinks.

What that does to the bill

Own the stack before the bill owns you.

Thirty minutes. Bring your current AI invoice. We look at what your teams send to model APIs, what it costs today, and what a self-hosted stack would cost instead. No deck, no discovery phase.

Book an assessment Email instead