AI Wisdom

Google Cloud professional level

Google Cloud Professional Data Engineer

Code
PDE
Length
120 minutes
Price
$200 plus tax
Valid for
2 years. A $100 renewal exam is available.
Study time, our estimate
80 to 110 hours

A strong, long-standing data credential that has picked up AI content, such as preparing unstructured data for embeddings and RAG. If your AI data sits in BigQuery, this is the exam to take.

Take it if

You build data platforms on Google Cloud.

Skip it if

Your data work is not on Google Cloud. Much of the exam is choosing between Google products.

Before you start

Check yourself against this list first. If something is missing, close that gap before you pay for an exam.

  • Three years in industryGoogle recommends this, with one year on Google Cloud.
  • SQL and pipeline experienceBatch and streaming basics.

Next: PMLE

All exam details
Format
40 to 50 multiple-choice and multiple-select questions
Length
120 minutes
Price
$200 plus tax
Valid for
2 years. A $100 renewal exam is available.
Languages
English, Japanese
Experience
3+ years in industry, 1+ year on Google Cloud
If you fail
Professional: wait 14, 60, then 365 days

What the exam covers

  • Designing data processing systems~22%
  • Ingesting and processing the data~25%
  • Storing the data~20%
  • Preparing and using data for analysis~15%
  • Maintaining and automating data workloads~18%

What to focus on

  • Choosing stores from access patterns: BigQuery, Bigtable, Spanner, AlloyDB, Cloud SQL.
  • Streaming with Pub/Sub and Dataflow: windowing and late-arriving data.
  • BigQuery design: partitioning, clustering, materialized views, editions and reservations.
  • Governance: IAM, CMEK, DLP, Dataplex (now documented as Knowledge Catalog) and data residency.
  • Orchestrating with Cloud Composer, and running pipelines through CI/CD.
  • Preparing unstructured data for embeddings and RAG.
Everything on one mapEvery topic and key idea in this certification.
mindmap
  root((PDE))
    Design
      Governance
      Residency
      Migration
    Ingest and process
      Streaming
      Spark
      Automation
    Store
      Bigtable
      Spanner
      Data platform
    Analyze and automate
      Partition and cluster
      Materialized views
      AI data
Which database?Start from the access pattern.
flowchart TD
  A{"How is data read?"} -->|"Analytics"| B["BigQuery"]
  A -->|"Fast key lookups at scale"| C["Bigtable"]
  A -->|"Global relational"| D["Spanner"]
  A -->|"Regional relational"| E["Cloud SQL or AlloyDB"]
  A -->|"Files and objects"| F["Cloud Storage"]

Lesson 1 of 4, ~22% of the exam

Design

Security, reliability and migration are decided here.

Governance
IAM, keys, and handling personal data.
Residency
Regional rules shape where data can live.
Migration
Know the transfer and replication tools.
Protecting a sensitive columnDifferent methods keep different abilities.
flowchart TD
  A{"What must still work?"} -->|"Joins"| B["Deterministic tokenization"]
  A -->|"Nothing, just hide it"| C["Masking"]
  A -->|"Reversal by a few people"| D["Encryption with managed keys"]

Common trap

Choosing a design before settling compliance.

Try it yourself

Mask a column with Sensitive Data Protection.

Official docs: Sensitive Data Protection

Lesson 2 of 4, ~25% of the exam

Ingest and process

The largest domain.

Streaming
Pub/Sub ingests, and Dataflow handles windows and late data.
Spark
Dataproc runs existing jobs with little change.
Automation
Cloud Composer runs dependent workflows.
A streaming pipelineWindows group events by when they happened.
flowchart LR
  A["Events"] --> B["Pub/Sub"]
  B --> C["Dataflow windows"]
  C --> D["BigQuery"]
  E["Late event"] -.->|"Allowed lateness"| C

Common trap

Rewriting Spark jobs when moving to Dataproc would do.

Try it yourself

Stream into BigQuery with a late data rule.

Official docs: Dataflow streaming pipelinesCloud Composer overview

Lesson 3 of 4, ~20% of the exam

Store

Choosing the store sets everything else.

Bigtable
High-throughput key lookups.
Spanner
Relational with global consistency.
Data platform
Google now documents Dataplex as Knowledge Catalog. Check the guide for the name it uses.
Choosing storage from access patternsRead this map from the workload outward.
mindmap
  root((Storage))
    Analytics
      BigQuery
    Key lookups
      Bigtable
    Relational
      Spanner global
      Cloud SQL regional
      AlloyDB
    Objects
      Cloud Storage
    Cache
      Memorystore

Common trap

Using BigQuery for millisecond key lookups.

Try it yourself

Write down when you would choose each database.

Official docs: BigtableSpannerKnowledge Catalog

Lesson 4 of 4, ~15% + ~18% of the exam

Analyze and automate

Covers AI data preparation and running it.

Partition and cluster
Reduce what queries scan.
Materialized views
Precompute heavy aggregations.
AI data
Prepare unstructured data for embeddings and RAG.
Partition, then clusterEach step reduces what is scanned.
flowchart LR
  A["Whole table"] -->|"Partition by date"| B["One day"]
  B -->|"Cluster by customer"| C["Few blocks"]
  C --> D["Lower cost"]

Common trap

Scanning a whole table when queries filter by date.

Try it yourself

Partition a large table and compare scanned bytes.

Official docs: Partitioned tablesClustered tablesMaterialized views

Study plan

  1. Weeks 1 and 2: pipelines

    Stream through Pub/Sub into Dataflow and BigQuery, with windowing and a late data rule.

  2. Week 3: storage

    Write down when you would choose each database, then partition and cluster a large BigQuery table and compare costs.

  3. Week 4: govern and operate

    Mask a column with DLP, set up Dataplex, and write a Composer DAG with retries.

  4. Exam week

    Do the sample questions and the mock below.

Exam day and after

  • Google renamed Vertex AI to Gemini Enterprise Agent Platform, and its exams were updated to match. Old notes use the former names, so check the exam guide for the current ones.
  • The certification pages do not state a pass mark for these exams.
  • Sample questions are free on each certification page, with no limit on attempts.
  • Google says this exam will soon update for branding changes, so check product names in the guide before you sit it.
Test yourself8 practice questions, scored by exam topic.
Start the mock exam

Keep going

Exam facts checked on October 31, 2026 against Professional Data Engineer exam guide. Lessons are our own explanations, so use the official docs as the final word, and confirm price and dates before booking.