Lead Databricks Data Engineer

A typical engagement publishes multi-dimensional financial data (for example a management P&L with entity, department, project, product, version, period and account dimensions) from an enterprise planning system as a single, row-secured consumption table in Databricks, refreshed several times a day. The quality bar:

• Exact reconciliation to source control totals; no data loss.

• Full fidelity of hierarchies, alternate hierarchies, aliases and attributes.

• Zero consumer outage during refresh; a failed run never replaces the last good data.

• Row-level security that matches the source system's access rules.

• Sub-second response for agreed benchmark queries.

• Additive schema changes absorbed without code changes; full rebuild from Git.

Key responsibilities

Architecture and technical leadership

• Own the Databricks solution design and drive open decisions with stakeholders: landing zone, publish mechanism, orchestration pattern, compute, retention and recovery targets.

• Translate integration specifications into buildable designs, data contracts and a delivery plan; lead design reviews with integration architects and platform teams.

• Lead, mentor and review the work of 2–4 data engineers; set coding, testing and documentation standards.

• Raise risks early (data volume, security parity, BI connectivity) and propose options with trade-offs.

Platform setup and governance (Unity Catalog)

• Set up catalogs, schemas, storage credentials, external locations and volumes for landing zones and medallion layers across dev, test and prod.

• Define the grant model with account-level groups synced from Microsoft Entra ID; configure service principals for jobs and CI/CD.

• Work with Azure network teams on private endpoints for ADLS and Network Connectivity Configurations for serverless compute.

Ingestion and orchestration

• Build manifest-gated, file-arrival-triggered Databricks Workflows with task dependencies, task values, condition-task publish gates, queueing, retries, timeouts and notifications.

• Implement Bronze ingestion from delimited files with checksum and row-count verification against a JSON manifest.

• Guarantee idempotent reruns and safe handling of overlapping or late runs.

Data modelling and transformation (Silver / Gold)

• Build PySpark logic to validate and flatten ragged parent-child hierarchies, detecting cycles, orphans, multi-parent members and non-unit weights.

• Handle alternate hierarchies through configuration, as flattened tables or bridge tables.

• Pivot long-form attributes and aliases into columns; build a contract-driven Gold generator that absorbs additive schema changes automatically.

• Design wide, denormalized consumption tables, including leaf/consolidated grain flags to prevent double counting.

Data quality and reconciliation

• Implement blocking quality checks (contract, referential integrity, hierarchy, typing, enrichment, security) and exact reconciliation to source control totals.

• Build control tables (run log, check results, reconciliation, schema contract) and an operations dashboard.

Security

• Implement source-system security parity: map source users to Entra identities, resolve group and member-level rights into a security lookup table, and enforce them with a Unity Catalog row-filter function.

• Keep data and security consistent during publish; define break-glass access, ownership and audit monitoring.

• Advise BI teams on per-user connectivity (SSO / DirectQuery) so row-level security holds end to end.

Publishing and performance

• Implement zero-outage publishing with atomic Delta overwrites or blue/green swaps, plus rollback via time travel.

• Tune for sub-second queries: liquid clustering, data-skipping statistics, predictive optimization, Photon, SQL Warehouse sizing and row-filter cost; run benchmark tests with real access profiles.

DevOps, testing and operations

• Package everything as Databricks Asset Bundles with CI/CD (GitHub Actions or Azure DevOps) across dev, test and prod.

• Own the test strategy: unit tests, contract and fixture tests, integration, SIT reconciliation, security UAT and performance tests.

• Write the runbook (rerun, rollback, full rebuild, schema change) and hand over to production support.

Collaboration

• Agree file contracts, manifests and control totals with source-system administrators.

• Work with finance users to defi