Opening your project
Real company data. Verifiable proof. On its way…
An independent ProoV educational experience using BMW Group as a publicly documented SAP HANA customer (nominative reference). Scenarios and datasets are synthetic, for educational use only — not affiliated with, endorsed by, or sponsored by BMW Group AG or SAP SE. You play a database engineer at Vektor Data GmbH, a fictional consultancy, optimising the SAP HANA cluster behind ConnectedDrive's dealer dashboard: design the schema, rewrite slow SQL, benchmark the gain, and brief the CTO.
Just completed
A real case study built on ProoV, completed step by step and graded by ProoV.
Step into the seat of a database engineer at a consultancy optimising the SAP HANA cluster behind BMW ConnectedDrive's dealer dashboard. Meet the team, grasp the scale of enterprise concurrency, and get briefed on the four engineering tasks ahead.
Learn what makes SAP HANA fast: in-memory column storage, dictionary compression, and the architecture BMW relies on. Then meet the villain — the slow query that is killing the dealer dashboard's SLA.
Load the ConnectedDrive telematics traces with pandas, analyse column cardinality, then write a CREATE COLUMN TABLE DDL schema that applies DICTIONARY encoding to the low-cardinality columns and types all twelve fields correctly.
Profile the trace log to find the two costliest SQL patterns (FULL_TABLE_SCAN and SELECT *), then rewrite them to target the column store with a SAP HANA parallel hint and quantify the speedup.
Call the verdict before you see the data, then load the benchmark CSV to compute CPU, latency, memory, and QPH deltas, check whether the 200 ms dealer SLA holds at 100 concurrent users, and chart the before/after.
Study a worked example, then write a 300–500 word executive pitch to BMW's CTO that opens with a stability guarantee, cites a specific benchmark number, links to the Dealer SLA, and avoids unexplained jargon.
Review a sample evaluation and the model answer, see the skills you built, and submit your four engineering artifacts for AI evaluation.
Analyse column cardinality and write a CREATE COLUMN TABLE DDL that applies dictionary encoding to the right low-cardinality columns and types every field correctly for an in-memory store.
Profile a trace log to find the costliest patterns, rewrite slow SQL to target the column store with parallel hints, then compute CPU, latency, memory, and QPH deltas and test them against a real SLA.
Translate benchmark numbers into a 300–500 word pitch to BMW's CTO that opens with a stability guarantee, cites a specific figure, links to the Dealer SLA, and avoids unexplained jargon.