The specific signals an experienced data engineer hiring panel grades on during the eight-second scan.

• Summary names the pipeline + the volume

  'Owns the 8B-event/day ingestion pipeline' beats 'data engineer.' Volume is the data-engineering scale signal.

• Orchestration stack named precisely

  Airflow, Dagster, Prefect, Argo Workflows. Generic 'orchestration tools' parses as junior.

• Warehouse + dialect named

  Snowflake, BigQuery, Databricks, Redshift, ClickHouse. Specific products parse as tokens.

• SLA + cost-per-1M-rows quantified

  Pipeline SLA in minutes, cost per 1M rows. The metrics hiring panels grade on.

• Data-quality work surfaced

  Great Expectations, dbt tests, custom checks. DQ work distinguishes senior data engineers from junior.

• One streaming or CDC bullet (where applicable)

  Kafka, Debezium, Kinesis, Pub/Sub. Streaming bullets signal current-vintage data engineering.

1. 1

   Open with the pipeline + volume

   A senior data-engineering summary names the pipeline and volume: 'Data engineer at a Series C SaaS; owns the 8B-event/day ingestion pipeline.' Mid: 'Data engineer on the warehouse team; owns the 14 critical-gold-table dbt project (Snowflake + dbt + Airflow).' Entry: 'Recent grad; shipped an end-to-end data pipeline as a capstone — Kafka ingestion + dbt models + Metabase dashboards used by 4 PI labs.'

2. 2

   Quantify with volume + SLA + cost

   Volume (events/day, GB/hour, TB warehouse). SLA (p99 minutes from event-write to gold table). Cost (compute spend per year, cost per 1M rows). DQ (test count, violation rate, ownership coverage).

3. 3

   Name the orchestrator + warehouse

   Airflow + Snowflake + dbt is the common stack. Dagster + BigQuery + dbt is another. Databricks + Delta + Unity Catalog is the lakehouse stack. Name your primary stack precisely; the JD will match.

4. 4

   Surface data-quality discipline

   DQ work distinguishes senior data engineers from junior. Pattern that works: • Name the DQ framework (Great Expectations, dbt tests, Soda, custom checks). • Quantify (test count + violation-rate outcome). • Name the ownership pattern (test owner, on-call rotation, escalation).

   DQ without an outcome metric reads as feature-list.

5. 5

   Close with OSS / community / talk

   High-signal closing: merged PRs to dbt-core, Apache Iceberg, Spark, Airflow, Polars. Talks at DataCouncil, Coalesce, Spark Summit. Substantial open-source data tools with adoption.

Each follows the [verb] [object] [number] structure hiring managers grade against. Copy them as a starting point, swap in your own numbers, and read the annotation to understand why each one works.

• Streaming

  Built the merchant-events ingestion pipeline (Kafka → Spark Structured Streaming → Delta Lake on Databricks); SLA p99 stays under 4 min from event-write to gold-table availability across an 8B-event/day workload.

  Why it works: Names architecture across three layers, SLA metric, volume, and gold-table availability detail.

• Cost

  Cut Snowflake compute spend $480k/yr (-32%) through three interventions: partition pruning on event-stream tables, materializing the 12 most-queried dashboards into incremental dbt models, and right-sizing the warehouse from M to S for off-hours queries.

  Why it works: Absolute dollars, %, and three interventions. Cost work that lands in dollars is data-engineering-specific senior signal.

• Data quality

  Built 38 dbt tests across critical gold-tables (freshness, uniqueness, referential integrity, business-rule checks); freshness violations dropped 84% within 6 weeks; ownership documented per test with on-call rotation.

  Why it works: Test count, category breakdown, freshness outcome, ownership pattern.

• Architecture

  Migrated the merchant-events ingestion from a batch Airflow daily-DAG to Kafka + Spark Structured Streaming + Delta merge; downstream latency from event-write to dashboard fell from 14 hours to 4 minutes.

  Why it works: Names the migration (batch → streaming), architecture, and a user-facing latency outcome (14h → 4m).

• dbt

  Owns 220 dbt models across 4 marts (finance, growth, product, ops); model-execution time fell 38% via incremental materialization + clustering keys on the event-stream tables.

  Why it works: Names model count, mart structure, and execution-time improvement. dbt scale + outcome is senior signal.

• Lineage

  Built the data-lineage observability layer (OpenLineage + Marquez); every dbt model and Airflow task now surfaces upstream/downstream impact in the catalog (used by 38 weekly active analysts).

  Why it works: Names the tools (OpenLineage + Marquez), the surface (catalog), and the WAU metric. Lineage tooling at this scale is senior signal.

• Data contracts

  Authored the org's data-contract framework (Protobuf schemas + dbt-source freshness + Great Expectations checks); 24 critical contracts published; upstream-breaking changes caught at PR-time, not in production.

  Why it works: Names the framework (Protobuf + dbt-source + GE), contract count, and the shift-left outcome. Data contracts work is current-vintage data engineering.

• Migration

  Migrated the data warehouse from Redshift to Snowflake over 4 months; 380 tables migrated; query latency p95 fell from 12s to 1.8s and concurrent-user capacity tripled.

  Why it works: Names source/destination, duration, table count, and two outcome metrics.

• CDC

  Built the CDC ingestion from PostgreSQL to the lakehouse (Debezium + Kafka + Spark merge); 14 source tables onboarded; downstream freshness SLA improved from 1 hour to 90 seconds.

  Why it works: Names the CDC stack (Debezium), the source/destination, and the freshness outcome. CDC work is data-engineering-specific senior signal.

• Open Source

  Two merged PRs to dbt-labs/dbt-core — one extended the on-run-end hook for incremental-model freshness reporting; one closed a partial-failure-recovery bug in incremental materialization.

  Why it works: Named project, PR count, two technical descriptions.

• Mentorship

  Mentored 2 analytics engineers transitioning into data-engineering focus; both shipped sole-owner production pipelines within 6 months.

  Why it works: Names the transition (AE → DE), timeframe, and deliverable per mentee.

• Entry-level

  Built an end-to-end pipeline as a university capstone (Kafka → Spark → Postgres → Metabase) for the campus IoT lab; 4,200 hourly events ingested; dashboard used by 4 PI labs through finals.

  Why it works: For entry-level data engineering, this kind of E2E shipped project is high-leverage.

Same intent, two phrasings. Read why the right column lands on the keep-pile and the wrong column doesn't.

Patterns our writers see most often when reviewing data engineer resumes — each one disqualifies candidates faster than weak experience does.

• Mistake

  Generic 'ETL' opener without naming the warehouse or orchestrator.

  Fix

  Name Snowflake / BigQuery / Databricks and Airflow / Dagster / Prefect by exact product.

• Mistake

  Pipeline bullets without volume.

  Fix

  Surface volume per pipeline (events/day, GB/hour, TB warehouse). Volume is the data-engineering scale signal.

• Mistake

  Cost bullets without dollar amounts.

  Fix

  Quantify in absolute dollars per year. Data warehouse cost work is graded in $.

• Mistake

  DQ bullets without an outcome metric.

  Fix

  Pair test count with violation-rate outcome and ownership pattern.

• Mistake

  Listing both batch + streaming as equal expertise without honest depth.

  Fix

  Lean toward one. Senior data engineering hiring panels distinguish the two flavors.

• Mistake

  Two-page resume.

  Fix

  One page until at least eight years experience.

• Mistake

  No mention of dbt or modern transformations.

  Fix

  If you ship in dbt, name it. dbt is universally recognized in 2026 data-engineering hiring.

• Mistake

  Generic 'big data' tokens (Hadoop, HBase) without current-vintage context.

  Fix

  Modern data engineering moved past Hadoop. Name current-vintage tools (Spark on Databricks, Iceberg, Delta) unless your role specifically uses legacy stacks.

Strong verbs lead strong bullets. Replace generic openers (worked on, helped with, was responsible for) with the specific verb that matches what you actually did.

ATS pipelines weight your Skills section as a structured list. Include 15-25 of the items below if they match your experience — not soft skills.