Level 4 exam — Python · GUIDINEE.COM

Exam — fresh questions

12 questions to validate the level: Django, Flask/FastAPI APIs, pandas data science, scikit-learn machine learning and polite scraping.
Answer everything, then click Check: your score and explanations will appear.

4.1 · Recall

1. Why is Django called “batteries included”?

It contains a built-in spreadsheet It runs without a database It plugs into an external battery It provides admin, ORM, auth and security out of the box

Admin, ORM, migrations, auth: everything provided and coherent.

4.3 · Recall

2. What is the core structure of pandas?

The DataFrame, a rows-and-columns table A programmable network router A deep learning model A video codec

The DataFrame: the programmable spreadsheet.

4.4 · Recall

3. What do fit() and predict() do in scikit-learn?

fit() learns from examples, predict() predicts on new data fit() deletes data and predict() restores it fit() deploys and predict() bills fit() prints and predict() draws

Learn (fit) then predict (predict), evaluate on never-seen data.

4.1 · Application

4. Nabil changes his Vente model (new date field). What to run?

Delete the database and start over makemigrations then migrate to evolve the table Rewrite the template by hand Restart the computer

Migrations translate the model into database evolution.

4.2 · Application

5. Nabil's POST /ventes returns 200 after creation. Fix?

Delete the route Keep 200, the body is enough Return 201 (created), 400 if input is invalid Return 404 every time

Truthful statuses: 201 created, 400 invalid.

4.2 · Application

6. Where does Nabil's interactive documentation on /docs come from?

A Word document maintained by hand Type hints validated by Pydantic, generated as OpenAPI A robot guessing the code It does not exist, it is a legend

Hints give auto validation + living documentation.

4.3 · Application

7. Nabil receives an unknown one-million-row CSV. First reflex?

Loop with for over each row Analyse it without opening it Draw ten random charts Inspect (head/info/describe/isna) then aggregate vectorised

Inspect first, vectorise instead of looping.

4.4 · Application

8. Nabil's model scores 100% on train but 60% in production. Diagnosis?

He evaluated on train (memory): mandatory train/test split The test must be deleted Charts are missing The model is perfect, prod is wrong

Without a split, you measure memory — not learning.

4.5 · Application

9. Nabil wants to scrape public prices while staying polite. Conduct?

Copy by hand every Monday Hammer a thousand requests per second Read ToS + robots.txt, space requests, prefer the official API Scrape with no user-agent or logs

Allowed, spaced, logged — or abstain.

4.3 · Arbitration

10. Nabil hesitates between GPU deep learning and RandomForest for his clients table. Decision?

Both mixed at random Neither, everything by hand Deep learning, it looks more impressive in meetings RandomForest as baseline: 80% of tabular with no GPU

Simple baseline first; deep for image/text via pre-trained.

4.5 · Arbitration

11. Nabil's target site forbids collection in its ToS but data is visible. Decision?

Scrape anyway, since it is visible Abstain or negotiate access/API, and get validation from a lawyer Scrape at night to stay unseen Resell the collected data

Visible is not allowed: ToS, personal data, lawyer.

4.4 · Arbitration

12. Nabil's arrears model penalises a whole neighbourhood. Reflex?

Raise the threshold to hide the problem Deploy it fast, figures are figures Audit biases, keep the human as decider, document limits Silently delete the neighbourhood from data

The model learns prejudices: audit + human decider.

What's next?

Score ≥ 80%: move to Level 5 — pro Python (quality, async, Docker, CI/CD) and take your specializations to production.
Score 50-79%: revisit the failed sections (Django 4.1, API 4.2, pandas 4.3, ML 4.4, scraping 4.5), reread the explanations above, then retake the test.
Score < 50%: restart the level from 4.1 coding each model (MVT, API, groupby, baseline, polite scraper), then retry the exam with a fresh head.