A fost o discutie despre experienta si despre ceea ce fac ei. Un feedback la o problema pe care am primit-o inainte de interviu. Apoi am avut de prezentat un use case asemanator cu ceea ce lucreaza ei. Compania se ocupa cu forarea minereurilor.
Use case: "Avem un client care are nevoie de o masinarie folosita pentru forarea aurului din pamant. Clientul respectiv ofera o multitudine de documente in format pdf cu detallile tehnice, iar intrebarea ar fi, cum as implementa o solutie de AI pentru ai ajuta pe angajatii companiei sa extraga din documente cateva informatii date de intervievator: numele companiei, rata de extractie, cantitatea totala, consumul de energie zilnica. Solutia ar fi, luarea unui LLM preantrenat, Llama/ChatGPT si combinat cu RAG pentru a extrage datele relevante din documentele date."

Inainte de interviu am avut un mic task pe care a trebuit sa il rezolv in 3 ore:
"The dataset consists of orders. Every row contains following information:
- Item_number: an ID of the item bought.
- Creation_date: the date the order was created.
- Price: price of the item. All prices are in the same currency.
- Supplier: name of the company, the item was bought from
- Region: region of the buyer.

Prices of the same item from the same supplier might differ based on the buyer region. We want to calculate features for a unique combination of item number, supplier, and region.
Write a function that creates a dataset containing the following columns on top of the already existing columns:
- Previous price of the specific item number – supplier – region combination. Name the column ‘previous_price’.
- Rolling average price of the last 3 transactions of the specific item number – supplier – region combination, excluding the previous price. If less than 3 records are available for a specific row, use as many as available. Name the column ‘price_roll3’.
You can drop any rows that contain missing values after the transformation.
"