Data engineering internships: what to learn and build

Find out which dataset the intern owns

A source connector, a shared orders table and a dashboard are different assignments. Ask what the output represents and who depends on it. The useful question is not simply whether the role uses SQL, but whether you will define, test or maintain the data behind someone’s decision.

For a project, use fictional orders with an order ID, version, amount and event date. Define one current row per order, with the highest accepted version as the current value. Preserve raw arrivals separately. That explicit rule lets another person decide whether your output is correct.

Five arrivals should not become $350 of orders

The following feed contains a duplicate and a correction. The amounts are invented. Under the stated current-order rule, an updated amount replaces the previous amount; it is not an additional purchase.

• Adding every arrival gives $350. The current orders are A $100, B $70 and C $30: $200.
• Replaying the same feed should leave the result at $200. Receiving B version 1 after version 2 should not restore its old $50 value.
• Two different amounts for the same order and version are a conflict, not a reason to take whichever arrived last. Keep the disputed input for review under your documented rule.

Show the correction in the report, not just the database

Keep both event time and arrival time. C arrived today but belongs to yesterday’s activity. A report that only refreshes today can remain wrong even when the current-order table is correct. Rebuild or update the affected reporting period, then compare its total with the accepted records.

Your project should include the input feed, current-order output, a small daily summary and tests for replay, an older version and a conflicting version. State how refunds or deletions would be represented; do not silently invent a rule for events you have not modeled.

After building it, a résumé entry can describe versioned corrections, late-arrival handling and repeatable totals. Do not call it production-scale without real operational evidence. In the interview, walk through the exact arrival that would break a naive sum, then show the test that catches it.

Before you apply

Eligibility

Check the posting’s SQL, programming and enrollment requirements. A data title alone does not tell you whether the work is engineering or analysis.

Timing

Datadog typically posts next-year student roles in September and conducts most interviews September–March, continuing until filled.

Location

Datadog describes office-based early-career placements in a hybrid workplace. Check the listed office.

Pay

Confirm whether the placement is a summer internship or a longer co-op before calculating total pay.

Sources: Datadog early careers

Job titles to search

Data engineering intern

Move, transform and maintain data for other people or systems.

Analytics engineering intern

Build tested data models and consistent definitions for analysis.

Data platform intern

Develop shared infrastructure used by data producers and consumers.

Where to look

Search routes, not confirmed vacancies.

Compare the work and try an example

Use a role comparison or worked exercise to choose your preparation before tailoring an application.

• Internship skills: distinguish required, preferred and taught on the job
• Build an economic consulting work sample that another person can reproduce
• Citadel versus Citadel Securities internships: choose the employer and role