Question 1

Joins and window functions - INNER, LEFT, RIGHT, FULL OUTER, ROW_NUMBER(), RANK(), DENSE_RANK()

Accepted Answer

Joins: INNER = intersection only; LEFT = all left + matching right (NULL fill); RIGHT = mirror of LEFT; FULL OUTER = union of both. Why it matters: Join choice affects result cardinality and NULL handling—wrong join = wrong business logic (e.g., LEFT to preserve all customers even without orders). Window functions: ROW_NUMBER() = unique rank 1,2,3; RANK() = ties same rank, gaps after; DENSE_RANK() = ties same rank, no gaps....

Question 2

What GCP tools do you use?

Accepted Answer

**Typical stack**: GCS (raw/curated storage) + Dataflow (streaming/batch) + BigQuery (analytics) + Cloud Composer (orchestration) + Data Catalog (governance). **Ingest**: Pub/Sub, Transfer Service, Cloud Storage. **Processing**: Dataflow (Beam), Dataproc (Spark). **Analytics**: BigQuery, Looker. **Governance**: Data Catalog, DLP. **Why GCP**: Strong integration (BigQuery + GCS native); serverless options; good ML (Vertex AI). **Scalability**: All scale elastically....

Question 3

Do you interact directly with business users?

Accepted Answer

**Situation**: Data engineering deliverables must align with business decisions.

**Task**: Bridge technical implementation and business needs through direct engagement.

**Action**: Requirement gathering with concrete examples; iterative delivery and UAT; ad-hoc analysis; incident communication. Translate technical constraints (latency, cost) into business trade-offs....

Question 4

Explain clustering with a real-time example.

Accepted Answer

**Example**: Customer segmentation—K-means on (frequency, recency, spend). Segments: high-value loyal, at-risk, new.

**Why Clustering**: Unsupervised; discovers patterns without labels. K-means iteratively assigns to nearest centroid.

**Scale**: MinHash/LSH for large data. Normalize features; choose K via elbow/silhouette....

Question 5

How are new requirements added in Agile?

Accepted Answer

**Flow**: Product backlog (prioritized); new items added and ordered by value. Sprint planning—top items pulled. Backlog refinement—clarify, estimate. Mid-sprint: negotiate with PO; can add if capacity or swap. Backlog is dynamic. Large requests: split into stories, prioritize....

Question 6

How do you handle a ticket beyond story point duration?

Accepted Answer

**Situation**: Ticket estimated 3 points, taking 5.

**Action**: (1) Communicate early—flag when approaching limit; (2) Re-scope—split remaining; complete current scope; (3) Swarm—get help if blocked; (4) Retro—why estimate off?; (5) Update—adjust points; don't stretch. Deliver minimal scope; create follow-up....

Question 7

How do you resolve source issues?

Accepted Answer

**Process**: (1) Validate—reproduce; logs, sample; (2) Isolate—source vs pipeline; (3) Communicate—notify owners with evidence; (4) Workaround—filter bad, retry; (5) Track—ticket; monitor. Example: API 500—retry + alert; engage API team with request IDs....

Question 8

Share experience handling a P1 issue.

Accepted Answer

STAR: SITUATION—P1: broken nightly batch, dashboards down. TASK—Restore within SLA. ACTION—Acknowledged in 15 min; identified schema change in source; rollback consumer; fixed schema handling; added validation. RESULT—Resolved in 45 min; runbook created. COMMUNICATION: Status updates; stakeholder notification....

Ford Data Engineer Interview Questions

+2 More Questions with Expert Answers

Other Companies

Ford Data Engineer Interview Questions

+2 More Questions with Expert Answers

Other Companies