Author name: Abhay Krishnan

Day 17 closes sub-objective 1.6 with two topics. Cortex Analyst converts natural language to SQL via a semantic model. Snowflake ML adds forecasting, anomaly detection, and classification as SQL-callable classes. The fact tested most often as a multi-select trap: ML Models are first-class database objects, one of the 12 from Day 5. 🗣️ Plain-English First […]

Day 15 covered pulling your code into Snowflake with Snowpark. Day 16 flips the direction. With Cortex, the AI already lives inside Snowflake. You call it from a SQL worksheet like any other function. No Python, no API keys, no model deployment. The common trap on C03 mocks is the line “Cortex AI requires Snowpark

Week 3 opens with the AI/ML developer stack: Snowpark, Notebooks, and Streamlit-in-Snowflake as one integrated developer surface. Snowpark hosts non-SQL code that runs inside Snowflake, which is the answer to official Sample Q5 and the most-tested sentence in sub-objective 1.6. 🗣️ Plain-English First Term Plain meaning Snowpark A set of client libraries (Python, Java, Scala)

Week 2 covers the densest stretch of Domain 1. Days 8 to 13 covered warehouses (sizes, Gen 2, Snowpark-Optimized), multi-cluster behaviour, and scaling policies. They also covered micro-partitions, table types, Iceberg, Dynamic Tables, and the three view types. Today is a closed-book drill, a gotcha map for recurring exam patterns, and a 10-question Week 2

Snowflake has three view types: standard, materialized, and secure. The exam tests which one solves which problem, plus the edition and sharing rules attached to each type. 🗣️ Plain-English First Term Plain meaning Standard view A saved SELECT. No data stored. Every query runs the underlying SQL fresh. Materialized view (MV) A view whose result

Apache Iceberg tables are Snowflake’s open lakehouse format with native read and write. Dynamic Tables are declarative pipelines refreshed against a freshness target instead of a schedule. Today covers what each one is and the catalog modes Iceberg supports. The keyword TARGET_LAG beats both Streams+Tasks and Materialized Views for hands-off pipelines. 🗣️ Plain-English First Term

Snowflake has four table types with nearly identical DDL but very different storage and recovery behaviour: permanent, transient, temporary, and external. The transient drop scenario is the trap that catches most COF-C03 candidates on the exam. 🗣️ Plain-English First Term Plain meaning Time Travel The window during which you can query, clone, or UNDROP historical

Snowflake stores every table in immutable columnar files called micro-partitions, and skips the ones a query does not need. Today covers what micro-partitions are, how pruning works, and when a clustering key is worth the credits. 🗣️ Plain-English First Term Plain meaning Micro-partition One small immutable file (50–500 MB uncompressed) holding a chunk of a

Day 8 covered warehouse size. Day 9 covers cluster count. Resizing fixes one slow query. A multi-cluster warehouse fixes many queries queuing at the same time. The COF-C03 tests this distinction repeatedly. Three facts carry most of the marks: scale UP for complex queries, scale OUT for concurrency, and Enterprise Edition is the minimum for

Week 2 starts with compute, the part of Snowflake that shows up on every monthly invoice. A virtual warehouse is the MPP cluster that runs your SQL, and three knobs (size, generation, and warehouse type) decide what you pay and how fast it runs. 🗣️ Plain-English First Term Plain meaning Virtual warehouse The compute cluster