tiyuvta
עברית

Teach a model your work, when it pays off.

Sometimes a general AI model does not know your terms, your formats or your language well enough. First we check whether better instructions solve it. If they do not, we train the model on examples from your work and compare the result before and after.

Talk about your examples

Who this is for

  • The answers are close, but miss your terminology, your document formats or your tone.
  • You work in Hebrew or another language where general models are weaker, and need to know how well they really do.

What you receive

  • A set of tests built from your real work.
  • A comparison of the starting model and the improved one.
  • The trained model, ready to run, where training is approved and the license allows it.
  • Running instructions within the agreed scope.

How it runs

  1. Build the tests

    We set examples aside so the result is judged fairly.

  2. Try the simple fix first

    We compare better instructions with training on your examples.

  3. Check the full picture

    Quality, speed and cost are measured together before we choose.

Hebrew starts with your examples

We research Hebrew models and how they break text into small pieces called tokens. Fewer tokens do not guarantee better answers, so answer quality is tested on examples from your task. Model serving concepts

Measuring first can save the cost of training

Before recommending training, we measure the starting model on your task. Our research notes show how we tell a real gap from a measurement mistake.

Browse the research

Scope, handover and support

Your proposal sets out the work, the price, the start and handover dates, and what we need from you. Running costs, such as hardware or cloud charges, are listed separately. On-site days and ongoing support are agreed up front, and the engineers who build your system keep supporting it.

Questions about the work

How is the price decided?

It is based on the work involved, the systems we connect and the support you want. Your proposal separates our work from running costs such as hardware or cloud charges.

Who keeps supporting the system?

Before we start, we agree on support coverage, on-site work and how problems are handled.

Do I need to buy hardware first?

No. We start from your work and your budget, compare the options, and only then recommend a purchase or a cloud setup.

Talk about your examples