Model Choice
An AI model that fits the problem
Good enough for the problem. Cheap enough to keep running. We test it on your data before anyone calls it production. You see next month's cost before you decide.
Examples
The engineering problem, in a few shapes. A table, a text, and a result someone will call production.
- A table with a predictable pattern. Classification, regression, clustering, forecasting, and anomaly detection, where a large model would be slower, more expensive, or less accurate than a smaller one.
- Text that is not a table. Unstructured text, open-ended language, or a task that requires reasoning. That is when a language model is the tool.
- An evaluation before the name production. Real questions, on your files, before anyone calls the result production.
Technologies
The model is chosen on accuracy, cost, latency, and where the data sits. The evaluation is how that choice is checked.
- Language models. OpenAI, Claude, Gemini, and open-source options like Llama, Mistral, Gemma, and Qwen, chosen per task based on accuracy, cost, latency, and privacy needs.
- Classical ML. Scikit-learn and XGBoost for tabular problems where a structured model beats a language model on both quality and cost.
- Where the data sits. The OpenAI or Anthropic API under an agreement that does not train on your data, or an open-source model on your infrastructure or ours, with nothing sent to an external API. The technical plan locks that decision before we build.
- Evaluation. Real questions, on your files, before production. The technical plan states the monthly cost.
How we'd work on this
A technical situation
A language model is put on a table a smaller model would score better, cheaper, and faster. Or the result is called production before it has been run on the real files.
How we'd approach it
We use classical ML when the table is enough, and a language model when the text is not. An evaluation runs on your files first. The technical plan states the monthly cost, and where the data sits.
What you'd get
You leave with that model running on your files, the monthly cost in the technical plan, the code, and you decide whether to continue.
Questions about model choice
LLMs when the input is unstructured text, open-ended language, or the task requires reasoning. Classical ML (XGBoost, scikit-learn) when the data is tabular and the pattern is predictable: churn prediction, sales forecasting, fraud detection. Classical ML tends to be cheaper, faster, and more accurate for those jobs.
That depends on the choice. When the case allows, we use the OpenAI or Anthropic API under agreements that prevent training on your data. When privacy requires it, we run open-source models like Llama, Mistral, Gemma, or Qwen on your infrastructure or ours, with no data sent to external APIs. The technical plan locks that decision before we build.
What you get
- A working fix on your data
- You keep the code
- 2 to 5 weeks when the owner and the files are available
Let's talk about your case
Talk to the Lab
Tell us the problem in a few lines. What the team does today, and where it breaks. We reply the same business day.
What happens next
- We reply the same business day
- Diagnostic in 3-5 days
- Working fix in 1-3 weeks. Technical plan in the last days, with the work done.
Start here