SnowPro Advanced: Data Scientist (DSA-C01)
Design, develop, and deploy scalable machine learning, feature engineering, and statistical modeling workflows in the Snowflake Data Cloud using Snowpark ML, Cortex AI, and Python.
Official Exam Specification & Quick Facts
Question Difficulty Breakdown
Exam Guidelines & Simulation Rules
Frequently Asked Questions (FAQ)
The DSA-C01 exam features 65 scenario, Python/Snowpark ML, and data science questions over 115 minutes. Cost is $375 USD with a passing threshold of 750/1000 (75%). SnowPro Core is a prerequisite.
Domains include: (1) Data Exploration and Preparation (~25-30%), (2) Feature Engineering with Snowpark (~25-30%), (3) Model Development and Training (~25-30%), and (4) Model Deployment and MLOps (~15-20%).
Snowpark ML provides Python APIs for preprocessing (scikit-learn and XGBoost transformers) and model training that execute pushdown operations directly inside Snowflake's virtual warehouses.
The Model Registry catalogs, versions, and manages trained machine learning models (scikit-learn, PyTorch, Hugging Face), tracking metrics and metadata, and enabling automated SQL inference.
Cortex ML Functions are pre-built, automated machine learning models exposed as SQL functions for Time Series Forecasting, Anomaly Detection, Contribution Analysis, and Top Insights.
Snowflake integrates an Anaconda channel directly into the platform, allowing data scientists to import curated, secure Python libraries (NumPy, pandas, scikit-learn, PyTorch) with zero installation.
Snowflake Cortex generates vector embeddings via `SNOWFLAKE.CORTEX.EMBED_TEXT_768()` and stores them in native `VECTOR` columns, queried using vector similarity distance functions (`VECTOR_COSINE_SIMILARITY`).
UDFs are designed for row-by-row or vectorized (pandas) transformations that return scalar values and can be called inside `SELECT` queries. Stored procedures perform administrative, orchestrating, or training tasks.
Vectorized UDFs accept batches of rows as pandas Series or DataFrames, enabling high-performance vectorized execution using SIMD and native C-extensions instead of row-by-row processing.
The Feature Store centralizes feature definitions, computes features continuously, and serves features consistently for model training and real-time prediction, eliminating train-serve skew.
The DSA-C01 credential is valid for two years from the pass date.
DSA-C01 certifies Senior Data Scientists, Machine Learning Engineers, and AI Researchers earning average US salaries between $145,000 and $190,000+.