$50–60/hr · micro1
You test how well AI assistants handle real analytics tasks by running workflows, verifying results, and scoring performance.
What you would do
- Execute structured evaluation scenarios that mirror real analytics work: anomaly detection, KPI reporting, data refreshes using AI-powered tools
- Write SQL queries to independently verify AI-generated figures and assess accuracy of joins, filters, and time window logic
- Maintain and reset Snowflake datasets to ensure clean test states and correct answers for each scenario
- Oversee access controls, roles, and permissions within the warehouse to keep evaluation environments secure and repeatable
- Configure and document integration steps across multiple tools to enable consistent, traceable testing
Who they want
- 3+ years of hands-on data analytics or analytics engineering work
- Advanced SQL skills and deep Snowflake expertise including warehouses, access management, and query history
- Demonstrated ability to audit and reconcile metrics against source data; keen eye for catching aggregation or logic errors
- Familiarity with AI tools like Claude and ChatGPT, with critical perspective on AI-generated SQL correctness
- Knowledge of SaaS platforms (Slack, Google Workspace, Microsoft 365) and experience with rubric-based evaluation or QA work
Main skills
What the interview asks about
1.Detecting AI SQL mistakes
AI can write syntactically correct SQL that produces subtly wrong results; independent verification is your job.
For example: “An AI assistant writes a query to find daily revenue outliers in the last 90 days, but you notice it groups by date without accounting for different sales regions. How do you verify the error and document it?”
2.Managing seeded test data integrity
If test data gets corrupted or mixed across runs, you can't trust any evaluation result; reset procedures matter as much as the test itself.
For example: “You're setting up 15 test scenarios in Snowflake and discover the seed data script doesn't track which version was loaded. How do you ensure each test starts from a clean, documented state?”
3.Reconciling reported metrics to source
A metric that looks plausible can hide a wrong join or timestamp filter; you must rebuild calculations to confirm.
For example: “An AI report says 22% of customers are in Chicago, but your customer base dashboard shows 18%. Walk me through your reconciliation process.”
4.Explaining access control decisions
Secure test environments require thoughtful permission design; you must document why each role grants what.
For example: “You're setting up read-only access for AI evaluations but need to exclude a few sensitive customer columns. How do you configure this in Snowflake and document it?”
5.Scoring consistency across evaluations
If you score harshly on one task and leniently on another, evaluation data becomes unreliable; calibration fixes this.
For example: “In a calibration session, you score a KPI report 7/10 for missing a region, but a peer scores the same error 5/10. How do you discuss and resolve this?”
A task you may get
Design a test scenario for AI: write the scenario description, seed data, and the SQL you'd run to verify AI output. Then score a sample AI response against a rubric you define.
How to prepare
- Review a sample KPI dashboard or analytics report and identify 5 common errors (e.g., time zone misalignment, duplicate rows from joins)
- Practice reconciling a metric: start with a dashboard number and rebuild it from raw tables to understand where errors hide
- Research Snowflake role and grant structures; understand how to create read-only access with column-level masking
The facts
- Pay
- $50–60/hr
- Open to
- Bangladesh, Hong Kong, India, Indonesia, Japan, Kazakhstan, Kyrgyzstan, Malaysia, Pakistan, Philippines, Singapore, Sri Lanka, Taiwan, Thailand, Uzbekistan, Vietnam, Austria, Belarus, Belgium, Denmark, France, Germany, Greece, Italy, Netherlands, Portugal, Russia, Spain, Switzerland, United Kingdom, Argentina, Brazil, Chile, Colombia, Mexico, Peru, Algeria, Bahrain, Egypt, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Palestine, Qatar, Saudi Arabia, Tunisia, United Arab Emirates, United States, Canada, Nigeria, Kenya, South Africa, Ghana, Ethiopia
- Field
- Software Engineering
- Role type
- Generalist
- Posted
- 8/18/2026
- Places left
- 4
We wrote this page from the public micro1 listing. It may be out of date, so read the full posting before you apply.