Snowflake QA and Reporting

Organize a multi-step analytical query with common table expressions.

Use ROW_NUMBER to select one required record from each business entity or group.

Calculate period-over-period changes with the LAG window function.

Prevent double counting by checking table grain and join cardinality before aggregation.

Translate a plain-language KPI definition into a SQL calculation at the required grain.

Compare independently calculated and reported figures with an exception query that exposes mismatches.

Detect missing, duplicate, unmatched, and out-of-range records with targeted SQL checks.

Investigate an anomalous KPI by segmenting it across relevant dimensions and time periods.

Set up a Snowflake trial account and execute a validation query in a Snowsight worksheet.

Explain how Snowflake separates storage, compute, and cloud services in an analytical workflow.

Create and navigate databases, schemas, tables, and views using fully qualified object names.

Create, start, suspend, resize, and select a virtual warehouse for a query workload.

Load a supplied CSV dataset into Snowflake using a stage, file format, and COPY command.

Validate a data load by checking row counts, column types, COPY history, and rejected records.

Use Snowsight query history to identify query status, duration, warehouse, and scanned data.

Control compute usage with warehouse auto-suspend, auto-resume, and statement-timeout settings.

Distinguish the functions of Snowflake users, roles, privileges, warehouses, databases, and schemas.

Create a functional role and grant only the privileges required for a specified analytical task.

Verify role inheritance and allowed or denied object access by switching roles.

Create a zero-copy clone of a schema for an isolated evaluation scenario.

Recover or inspect changed data with Snowflake Time Travel.

Reset a seeded evaluation dataset to a documented baseline using a clone-based procedure.

Apply a controlled DML change inside a transaction and choose whether to commit or roll it back.

Produce a role-and-privilege matrix documenting who can access each evaluation resource.

Translate an analytical requirement into explicit acceptance criteria and expected results.

Design SQL checks for completeness, uniqueness, validity, consistency, and referential integrity.

Verify an AI-generated join by checking join keys, cardinality, unmatched rows, and duplicate amplification.

Verify an AI-generated filter by testing included, excluded, NULL, and boundary records.

Verify a time-window calculation by checking date boundaries, time zones, and comparison periods.

Reconcile an AI-generated aggregate against an independently calculated source-of-truth query.

Create a compact seeded dataset containing edge cases that expose common analytical errors.

Write a reproducible defect report containing evidence, impact, expected behavior, and actual behavior.

Define a KPI with an owner, grain, formula, filters, time window, and refresh expectation.

Create a Snowflake reporting view that exposes consistent dimensions and measures.

Configure a Power BI Desktop connection to Snowflake using the correct account, warehouse, database, role, and authentication settings.

Choose between Import and DirectQuery based on freshness, performance, and governance requirements.

Build a compact Power BI semantic model with explicit table relationships and a date table.

Create DAX measures for a total, rate, and period-over-period change.

Validate Power BI visual totals against independent Snowflake SQL results.

Diagnose a deliberately stale or failed report refresh using connection settings, refresh evidence, and Snowflake query history.

Decompose an AI-generated analytical response into factual claims that can be tested independently.

Create a verification matrix that maps each claim to its business definition, test method, expected result, and required evidence.

Create a scoring rubric that separates numerical accuracy, SQL logic, explanation quality, and reproducibility.

Score an AI-generated result and support every judgment with query evidence.

Use single-variable experiments to isolate whether unfamiliar behavior originates in the data, SQL, role, warehouse, or client configuration.

Create a reproducibility record containing inputs, environment, role, warehouse, steps, and observed output.

Resolve two conflicting evaluation scores by applying shared rubric criteria to the same evidence and documenting the final decision.

Write a concise connectivity and authentication runbook that excludes secrets and sensitive credentials.

Create a Git repository with a professional structure for SQL, Power BI files, screenshots, data documentation, and the project README.

Load and validate a compact seeded sales or operations dataset in Snowflake.

Write unambiguous specifications for three portfolio KPIs.

Create and test a small Snowflake reporting layer that calculates the three KPIs at their required grains.

Investigate one seeded KPI anomaly and identify its contributing segment or data-quality issue.

Build a focused Power BI report page that communicates KPI trends and the anomaly.

Reconcile every displayed KPI with independent Snowflake verification queries and preserve the evidence.

Publish a project case study describing the business question, architecture, QA checks, findings, and limitations.

Design three realistic AI-assistant evaluation tasks that collectively test joins, filters, time windows, KPIs, and anomaly reasoning.

Seed a Snowflake test dataset with duplicate, missing, boundary-date, and unmatched-key edge cases.

Write a source-of-truth SQL answer and expected result for each evaluation task.

Capture candidate AI-generated SQL, prompts, execution results, and discrepancies in a standardized evidence template.

Apply a weighted rubric to score correctness, reasoning quality, and reproducibility for each candidate answer.

Write evidence-based defect reports for at least two flawed outputs, introducing controlled flaws if required.

Create and verify a reset script that returns the benchmark dataset to its baseline state.

Publish a benchmark report summarizing failure patterns, severity, evidence, and recommended safeguards.

Create environment-owner, evaluator, and read-only roles using least-privilege grants.

Configure a baseline schema, an isolated evaluation-run schema, and an appropriately sized virtual warehouse.

Implement and verify a clone-based procedure for resetting an evaluation scenario.

Test allowed and denied operations for every role and record the results in an access-control matrix.

Configure and document a Snowflake-to-Power BI connection without exposing credentials.

Run a controlled reporting refresh and verify the refreshed figures against Snowflake.

Produce an operational runbook with a compact architecture diagram covering access, reset, refresh, and troubleshooting procedures.

Publish a master portfolio overview and write truthful, impact-focused CV bullet points for all three projects.