Google Cloud Professional Data Engineer (PDE) Study Plan for 2026

Topics

• • PDE exam guide and domain mapping
• • GCP data services overview
• • BigQuery fundamentals and query model
• • Apache Beam programming model basics
• • Batch vs streaming architecture patterns

Activities

• • Read the official PDE exam guide end to end. Highlight the domain you know least about.
• • Set up a GCP project with BigQuery, Dataflow, Pub/Sub, and Cloud Composer APIs enabled.
• • Run a BigQuery query on a public dataset (e.g., `bigquery-public-data.github_repos.commits`) using the bq CLI.
• • Take an ExamOS Mock Exam in Rookie mode (30 questions). Note any domain where you scored below 60%.

Topics

• • BigQuery partitioned and clustered tables
• • BigQuery external tables and BigLake
• • Cloud Storage and file format selection
• • Bigtable row key design and hotspots
• • Storage service selection framework

Activities

• • Create a BigQuery table with date partitioning and clustering. Query `INFORMATION_SCHEMA.JOBS` to compare scan bytes against an unpartitioned copy of the same data.
• • Load Parquet data from Cloud Storage into BigQuery. Compare load time and query performance against the same data in CSV format.
• • Design a Bigtable row key for an IoT time-series workload that distributes writes evenly across nodes. Verify with the Key Visualizer.
• • Take an ExamOS Daily Sprint in Challenger mode. Repeat until you pass 2 times in a row.

Topics

• • Batch pipeline architecture patterns
• • Streaming pipeline architecture
• • Lambda vs kappa architecture
• • Pub/Sub design and configuration
• • Event-driven architecture patterns
• • Pipeline orchestration selection

Activities

• • Design a complete batch pipeline on paper: Cloud Storage to Dataflow (or Dataproc) to BigQuery. Write down the service selection rationale for each component.
• • Design a streaming pipeline: Pub/Sub to Dataflow with windowed aggregation to BigQuery. Document the windowing and trigger choices.
• • Create a Pub/Sub topic with ordering keys and a dead-letter topic. Publish test messages and verify ordering.
• • Take an ExamOS Daily Sprint in Challenger mode. Repeat until you pass 2 times in a row.

Topics

• • Apache Beam programming model and transforms
• • Windowing strategies and triggers
• • Dataflow templates and deployment
• • Dataflow autoscaling and streaming engine
• • Error handling and dead-letter patterns

Activities

• • Write a Dataflow pipeline with fixed windowing that counts events per minute from a Pub/Sub topic. Deploy it and verify output in BigQuery.
• • Implement a dead-letter pattern for unparseable messages: route failed elements to a separate Pub/Sub topic.
• • Deploy a Dataflow flex template from a Python script. Verify the job runs and scales.
• • Run the same input through fixed vs session windowing. Write a comparison explaining when each is the correct choice.
• • Take an ExamOS Daily Sprint in Challenger mode. Repeat until you pass 2 times in a row.

Topics

• • Dataproc cluster management and serverless
• • Spark on Dataproc and job submission
• • BigQuery loading methods and Storage Write API
• • BigQuery DML and MERGE statements
• • Cloud Composer DAG design and operators
• • Datastream for change data capture

Activities

• • Create an ephemeral Dataproc cluster, submit a PySpark word-count job, and delete the cluster. Document the full lifecycle.
• • Load data into BigQuery using three methods: batch load, streaming inserts, and Storage Write API. Compare latency and cost for each.
• • Write a BigQuery MERGE statement that upserts records from a staging table into a production table. Verify idempotency by running it twice.
• • Build a Cloud Composer DAG that triggers a Dataproc job and loads the output into BigQuery. Verify the full orchestration end to end.
• • Take an ExamOS Daily Sprint in Challenger mode. Repeat until you pass 2 times in a row.

Topics

• • BigQuery ML model training and prediction
• • Vertex AI model deployment and serving
• • Vertex AI Feature Store design
• • Dataflow and BigQuery monitoring and alerting
• • BigQuery performance optimization and cost
• • BigQuery reservations and slot management

Activities

• • Train a BQML logistic regression model on a public dataset. Evaluate it with `ML.EVALUATE` and interpret precision, recall, and AUC.
• • Create a Vertex AI endpoint and deploy a pre-trained model. Send a prediction request via the REST API.
• • Create a Cloud Monitoring alert on Dataflow system lag exceeding 30 seconds. Verify it triggers when lag spikes.
• • Query `INFORMATION_SCHEMA.JOBS_BY_PROJECT` to find the top 5 most expensive queries in the last 7 days. Identify one query you can optimize.
• • Take an ExamOS Daily Sprint in Challenger mode. Repeat until you pass 2 times in a row.

Topics

• • Dataplex lakes, zones, and metadata discovery
• • Dataplex data quality rules and scans
• • Data lineage and data catalog
• • BigQuery access control and row-level security
• • Cloud DLP de-identification transforms
• • CMEK, VPC Service Controls, and audit logging

Activities

• • Create a Dataplex lake with a curated zone and attach a BigQuery dataset. Run metadata discovery and verify the auto-generated entries.
• • Configure a Dataplex data quality scan with completeness and uniqueness rules on a sample table. Review the scan results.
• • Implement column-level security on a BigQuery table, restricting access to a PII column for a specific IAM group.
• • Create an authorized view that exposes aggregated data without revealing underlying PII rows. Test access with a non-admin account.
• • Configure a Cloud DLP inspection job scanning a Cloud Storage bucket for email addresses and phone numbers. Review the findings.
• • Take an ExamOS Daily Sprint in Challenger mode. Repeat until you pass 2 times in a row.

Topics

• • Full syllabus review across all five domains
• • Time management (120 minutes for 40-50 questions)
• • Service selection under time pressure
• • Multiple-select question technique

Activities

• • Take the official Google practice exam for PDE if available. Review every wrong answer.
• • Take at least 3 ExamOS Mock Exams in Legend mode. For every question you miss, write one sentence explaining the service selection principle behind the correct answer.
• • Simulate real exam conditions at least once: no phone, no breaks, 120 minutes, same time of day you plan to sit the exam.
• • Review your weak domains list from Week 1. For any domain still below confidence, do a focused Bonus Quest on that domain.