NCP-ADS Accelerated Data Science Professional Study Plan¶
8-Week Intensive Study Schedule¶
Phase 1: Foundation (Weeks 1-2)¶
Week 1: RAPIDS Fundamentals and cuDF¶
- Install RAPIDS environment (conda or Docker)
- Study RAPIDS ecosystem components
- Learn cuDF data loading (CSV, Parquet, JSON)
- Practice DataFrame operations (filter, groupby, merge)
- Understand pandas vs cuDF differences
- Reference: RAPIDS Docs
Week 2: cuML Machine Learning¶
- Study cuML algorithm categories
- Practice classification (Random Forest, Logistic Regression)
- Practice clustering (K-Means, DBSCAN)
- Learn dimensionality reduction (PCA, UMAP)
- Study model evaluation and cross-validation
- Reference: cuML Docs
Phase 2: Advanced Topics (Weeks 3-5)¶
Week 3: cuGraph and Graph Analytics¶
- Study graph construction from cuDF edge lists
- Practice PageRank, BFS, SSSP algorithms
- Learn community detection (Louvain, Leiden)
- Integrate cuGraph results with cuDF analysis
- Reference: cuGraph Docs
Week 4: GPU ETL and Dask-cuDF¶
- Build GPU-accelerated ETL pipelines with cuDF
- Set up Dask-cuDF for multi-GPU processing
- Practice scaling from single to multi-GPU
- Study memory management and RMM
Week 5: Spark Integration¶
- Configure RAPIDS Accelerator for Spark
- Run Spark queries with GPU acceleration
- Study supported vs unsupported operations
- Practice performance tuning
- Reference: Spark RAPIDS Docs
Phase 3: Review and Exam Prep (Weeks 6-8)¶
- Work through scenario-based questions
- Review all domain areas
- Focus on API knowledge and code patterns
- Full-length practice sessions
Self-Assessment Questions¶
- Can I write cuDF code for common DataFrame operations?
- Do I know which cuML algorithms are available and their API?
- Can I build and analyze a graph with cuGraph?
- Do I understand Dask-cuDF scaling patterns?
- Can I configure Spark RAPIDS for GPU acceleration?