Python

How many clients need to be processed at once, and how to determine the required capacity (for example, with a database of 5 billion clients)?

Am I correct in understanding that you first looked for the minimal match (for example, paragraph match), and then simply unloaded the entire document and sent it to the LLM context?

How will this model/service be launched and operate, considering that SMS needs to be sent at a specific scheduled time?

If this is an API, what will be in the body of the incoming HTTP request, and what does the model return?

So, it is necessary to monitor not only the quality metrics of the model but also the data on which this model is built, correct?

A simpler example: an SMS campaign offering to buy sausage, but a Christian fast is approaching — what changes in this situation and what should be paid attention to?

The bank [name] wants to send SMS about a new mortgage offer not to all clients, but only to targeted ones — what model is needed here and what is important to consider from a business perspective?

We'll wrap the model in a container, right?

What exactly does Kubernetes check before starting a service (readiness)?

What ML models, besides LLM, do you know from a business application perspective?

What is your experience with Kubernetes and Docker?

In your opinion, are all statistical quality metrics considered the same, or do they have nuances?

What are the other statistical quality metrics of a model — what are they?

The model service is constantly hanging, and clients are sent to it once a week — doesn't this seem inefficient, and what can be done?

How do we determine the required infrastructure capacity for a model?

What happens to data after the model evaluates it?

How exactly will the model be deployed — for example, on a virtual machine? Where specifically?

What else needs to be checked/monitored when running a batch job, besides meeting the deadline?

The Central Bank rate changes, and accordingly the credit conditions (amount, rate) change — what can change in the model/campaign as a result?

How does the model interact with data when it is already in production?