Senior Data Scientist - Chalhoub Group

What is your current monthly salary including currency?

Gender

We would like to know more about our candidates. Please note that any of the information that you provide does not affect your chances of being awarded the role. If you are happy to, please provide us with the information below. This information is completely voluntary.

Male

Female

Prefer not to Say

Are you or any of your family members currently working with Chalhoub Group or any of our Joint Venture Partners?

Yes - Currently working with Chalhoub Group

Yes - Currently working with a Joint Venture

Yes - I have a family member currently working with Chalhoub Group

Yes - I previously worked for the Chalhoub Group or a Joint Venture partner

None of the above

Where did you study?

Are you a UAE National? *

Yes

Do you hold a valid Emirati Family book (Khulasat Alqaid)?

Yes

You are to build a model to predict rare customer churn, positive class rate 5 percent. Which metric is most appropriate to use during model selection?

Accuracy

ROC AUC

Precision-Recall AUC

Mean Squared Error

You have a categorical feature with 10,000 distinct values to use in a tree-based model for production. Which encoding is most practical and safe?

One-hot encoding

Alphabetical label encoding

Target or frequency encoding with regularization and out-of-fold smoothing

Standard scaling

You deployed a scoring model where true labels arrive after 30 to 90 days. Which monitoring and retraining strategy best balances early detection and safe updates?

Monitor delayed label performance and retrain when it drops below a threshold.

Monitor input distributions with PSI and retrain monthly on a fixed schedule.

Wait for business users to report issues, then investigate and retrain as needed.

Combine delayed-label checks with near-real-time proxies, alert on multi-signal degradation, and use controlled rollouts such as champion-challenger before full swap.

You must deploy a fraud model across multiple business units with different schemas and constraints. Select all that apply.

Use a single core model with unit-specific preprocessing adapters, centralized experiment tracking, schema validation, and CI/CD with per-unit canary testing.

Deploy the identical containerized artifact everywhere with no local adaptation.

Standardize inputs via a feature store, implement automated monitoring and retrain triggers based on drift or KPI changes.

Build separate models per unit but register them centrally and enforce consistent governance artifacts such as model cards, lineage, and bias reports.

You rolled out a personalized high-value offer model for loyalty customers. One customer segment receives significantly fewer offers and converts less. Select up to three actions you would take.

Run a subgroup audit: measure allocation, acceptance rate, uplift, and business KPIs by segment.

Retrain with targeted mitigation: reweight or augment under-served segment data, add group-aware features, validate with cross-group holdouts.

Remove sensitive attributes such as region or nationality from model inputs and redeploy.

Increase offer generosity for the under-served segment universally until acceptance matches others, then stop monitoring.

Implement allocation guardrails: set minimum exposure or controlled exploration for the under-served segment, monitor short-term revenue and retention.

Personal information

First name

Last name

Phone

Upload CV

Additional files