SELECT, joins, aggregations, windows, CTEs, subqueries, indexes, execution plans.
Before starting, we run a 1-minute tech check — microphone, ambient noise, connection. If your setup isn't good enough, the test is fully refunded.
Show in 15 minutes that your SQL skills are the real deal — from multi-table joins to execution plans — with an AI oral exam that goes way beyond a LinkedIn checkbox.
The Plume SQL badge tests your ability to write, optimize, and explain SQL queries in real business contexts. During the 15-minute AI-led oral, you'll be asked to walk through advanced SELECT patterns, multi-table joins, aggregations, window functions (ROW_NUMBER, LAG, SUM OVER PARTITION BY), recursive CTEs, correlated subqueries, and execution plan analysis with EXPLAIN ANALYZE. This isn't a multiple-choice quiz: the AI asks you to narrate real situations, justify your join choices, reason about indexes, and place SQL in your broader data workflow.
What sets this badge apart from a self-declared skill is proof through conversation. Anyone can tick "SQL" on LinkedIn — you'll have a 0-to-100 score, a certified level (Novice / Proficient / Advanced / Expert), a detailed qualitative report written by Claude Opus, and a public link to your result. Hiring managers and data team leads can use it to compare candidates on a level playing field, without running their own time-consuming technical screens.
This badge is built for data analysts, data engineers, analytics engineers, and backend developers who live in SQL day-to-day — on PostgreSQL, BigQuery, Snowflake, DuckDB, or any other engine — and want to make that expertise visible and credible. It's equally valuable for career-switchers into data who need to prove their SQL chops before an interview, or for freelancers who want to give clients confidence in their technical rigor before signing a contract.
Here are the concrete dimensions the AI examines during the 15-minute oral.
Writing INNER, LEFT, RIGHT, and FULL OUTER JOINs as well as self-joins across 4 to 5 tables, explaining cardinalities, fan-out risks, and the difference between filtering with ON vs WHERE.
Mastery of ROW_NUMBER, RANK, DENSE_RANK, LAG, LEAD, FIRST_VALUE, and rolling aggregations with SUM OVER (PARTITION BY ... ORDER BY ...) to replace nested subqueries or awkward GROUP BY workarounds.
Using Common Table Expressions (WITH ... AS) to structure complex queries for readability vs performance, and distinguishing correlated from non-correlated subqueries and when each makes sense.
Reading EXPLAIN / EXPLAIN ANALYZE output, spotting expensive Seq Scans, choosing between B-tree, GIN, and composite indexes, and rewriting queries to cut estimated cost on large tables.
Using GROUP BY, HAVING, GROUPING SETS, ROLLUP, and CUBE to produce multi-dimensional reports, and handling NULL propagation correctly inside aggregate functions.
Positioning SQL logic inside a dbt project, an Airflow DAG, or a BI layer like Looker or Metabase: which transformations belong in SQL, how to version models, and how to test data quality downstream.
Knowing when to hand off to pandas, PySpark, or a vector store, with concrete examples of production cases where SQL hit its ceiling and a different tool was the right call.
Familiarity with engines like DuckDB, BigQuery, and Snowflake, syntax extensions like MERGE, recursive CTEs, and QUALIFY, plus the implications for partitioning, clustering, and materialized view strategies.
Final scoring is performed by Claude (Anthropic), which reads back the full transcript and applies this weighted criteria grid.
The queries and concepts described are syntactically correct and semantically suited to the business context. The candidate doesn't confuse JOIN with UNION, HAVING with WHERE, or clustered with non-clustered indexes.
Ability to diagnose a slow query using EXPLAIN ANALYZE, pick the right index type, avoid N+1 patterns, and anticipate how data volume affects query runtime across different engines.
The candidate articulates their thought process logically with concrete examples drawn from real experience. Technical terms are used precisely and defined when needed, without vague hand-waving.
Appropriate use of window functions, recursive CTEs, correlated subqueries, and set operators (INTERSECT, EXCEPT) in contexts where they genuinely improve readability or performance over simpler alternatives.
Knowledge of modern SQL engines (BigQuery, Snowflake, DuckDB), ability to place SQL in a full data stack, and solid judgment on when to move computation outside SQL to pandas, PySpark, or other tools.
A Plume session takes about 20 minutes, from tech check to badge delivery.
Before the exam starts, Plume verifies your mic and connection. No download, no account setup beyond the basics — the whole exam runs in the browser.
The AI asks you to introduce yourself briefly and describe the most complex SQL query you've written recently: the business context, the tables involved, and the technical choices you made.
The AI probes 3 to 5 themes: optimizing slow production queries, window functions vs GROUP BY, reading execution plans, integrating SQL in a dbt or Airflow pipeline, and knowing when to reach for PySpark instead. Questions adapt in real time to your answers.
The AI asks how you see SQL evolving in your domain — columnar engines, DuckDB, SQL-on-everything — and gives you a chance to mention anything you didn't get to cover during the deep-dive.
Claude Opus reads the full transcript and generates your score (0-100), your certified level, and a detailed qualitative report. Your badge is shareable immediately via a public URL you control.
Your score out of 100 translates into a level a recruiter can grasp at a glance.
You write basic SELECT statements with WHERE filters and simple JOINs, but struggle with GROUP BY plus HAVING, correlated subqueries, or reading an execution plan. You mostly work on single-table queries or small datasets.
You combine JOINs, GROUP BY, and subqueries without friction, you've started using CTEs to keep queries readable, and you can create a basic index. You spot slow queries but rely on documentation or a colleague for the deeper optimization work.
You reach for window functions (ROW_NUMBER, LAG, SUM OVER PARTITION BY) naturally, you read EXPLAIN ANALYZE output to choose your indexes, you structure SQL transformations in dbt, and you know when to hand off computation to PySpark or pandas.
You optimize critical queries against tables with billions of rows, you're comfortable with partitioning, clustering, and materialized views on BigQuery or Snowflake, you define SQL standards for your team, and you design data models with future queryability in mind from day one.
No degree or years of experience required to take the badge. Here are the profiles it makes the most sense for.
You spend your days in PostgreSQL, BigQuery, or Redshift and want to prove your SQL goes well beyond basic SELECTs — especially to recruiters who don't run their own technical screens.
You build pipelines with dbt or Spark SQL and want to certify your ability to write performant queries, design sound models, and diagnose production bottlenecks in a credible, shareable format.
You build Looker, Metabase, or Power BI models and want to show that the underlying SQL — aggregations, CTEs, optimization — is solid, not just the dashboards sitting on top of it.
You've learned SQL through a bootcamp or self-study and want an objective measure of your level before your next interview or freelance pitch, without sitting through a different test for every recruiter.
You write complex queries in your ORM or raw SQL every day and want to demonstrate you can talk performance, indexes, and execution plans at the same level as a dedicated data specialist.
Where and how your SQL badge will help you day to day.
You add your badge link to your LinkedIn profile before interviewing at a scale-up. The recruiter clicks through, sees your 78/100 score (Advanced level), and immediately understands you can handle complex queries — no separate technical test needed.
A client is choosing between two freelancers. You share your Plume SQL badge in your proposal. The detailed report shows your CTE and window function fluency plus BigQuery experience, tipping the decision in your favor.
You're going for a data lead role at your current company. Your Expert SQL badge gives your internal application an objective anchor that complements your project history and peer reviews.
At the end of a three-month data bootcamp, you take the badge to get an honest read on your actual level before applying for jobs. The report pinpoints exactly what to work on — optimization, indexes, execution plans.
You add your SQL badge link to your GitHub portfolio alongside your dbt models. Visitors can verify in one click that your command of declarative SQL and data testing is independently certified.
A data team receives 80 applications for an analyst role. By asking for a Plume SQL badge upfront, they cut screening time by 60% and focus human interviews on Advanced and Expert candidates only.
A few minutes to check you have everything you need.
At the end of your session you don't just get a score — here's everything that awaits you.
You get a score from 0 to 100 and a certified level (Novice, Proficient, Advanced, Expert) that accurately reflects your SQL depth — from basic joins to execution plan tuning on BigQuery or Snowflake.
Claude Opus writes a multi-paragraph report commenting on your strengths (e.g., CTE structuring, index reasoning) and specific areas to improve, based on what you actually said during the exam.
Your exam audio is stored securely and accessible only by you. Re-listen to prep for your next interview, or identify exactly where you hesitated on optimization or window function questions.
A unique public link lets you share your SQL badge on LinkedIn, in your resume, a freelance proposal, or your GitHub portfolio — readable at a glance by any recruiter or client.
Discover related skills you can validate with Plume.
A 15-min oral exam with an AI, a shareable badge for your recruiters.
Choose this badge · €19.99
