Skip to content

Services

What we take on.

Six services across four practices. Most engagements combine two or three of them, because the model is rarely the only thing that needs building.

machine learning

Models, and the systems that use them.

LLM applications

The hard part of an LLM project is usually not the model. It is how documents get in, how answers get checked, and what happens when the model is wrong. We build that whole path, including the review steps that stop a bad answer from reaching a customer.

What you get

  • Retrieval over your documents and databases
  • Agents and tool use wired into existing systems
  • Prompt, context, and cost engineering
  • Human review paths for high-stakes answers

Custom ML models

Not every problem needs a large general model. Some need a small one fit closely to your data. We start by checking whether your data can answer the question you are asking, and we tell you if it cannot. After that comes training, validation, and getting it into production.

What you get

  • Feasibility review before anyone writes training code
  • Feature and dataset construction from your systems
  • Training, validation, and error analysis
  • Deployment, monitoring, and a retraining plan

Tuning and evaluation

It is easy to change a prompt, look at one good output, and call it an improvement. We build evaluation sets out of your real traffic and score against them. Changes come with numbers attached, and regressions turn up before your customers find them.

What you get

  • Evaluation sets built from your real traffic
  • Automated scoring and regression runs
  • Fine-tuning where it beats prompting, and evidence when it does not
  • Latency and cost benchmarking

product

The application your users see.

Web and mobile applications

The software around a model still has to be good software. We take products from the first conversation to something running in front of real users. That covers the interface people use, the services behind it, and the release process that gets changes out safely.

What you get

  • Requirements and scoping you can read in one sitting
  • Interface design and front-end build
  • APIs, databases, and background processing
  • Release, monitoring, and on-call handover

infrastructure

What it runs on.

Cloud and infrastructure

We size infrastructure for the load you actually have, move existing systems onto it without a weekend of downtime, and define it in code so the next change is a pull request instead of a guess. Most engagements here start with a look at the bill.

What you get

  • Architecture review and cost reduction
  • Migrations from on-premise or another provider
  • Infrastructure as code, in your accounts
  • Backups, monitoring, and access control

advisory

Help with the calls that are hard to reverse.

Fractional CTO

Some decisions are expensive to reverse: what to build in-house, which vendor to commit to, who to hire first. We sit in on those, review what your team is building, and give you an answer with the reasoning behind it.

What you get

  • Standing time each month, not a support queue
  • Architecture and code review for your team
  • Vendor and build-versus-buy evaluations
  • Interviewing and technical hiring support

Not sure which of these you need?

That is a normal place to start. Describe what is not working and we will tell you which of the above applies, including when the answer is none of them.

Tell us what you are trying to build.

Send one email describing the problem. You will hear back from an engineer, usually the same day.

team@randomotter.comOther ways to reach us