Senior Data Platform Architect

End-to-end platform architecture and implementation

• Design the medallion architecture: catalog and schema design, layer contracts, and the promotion path from development through QA to production.


• Design robust, scalable data platform architectures aligned with enterprise security and networking standards.


• Act as a bridge between Data Solutions and Infrastructure, ensuring the platform decisions you make are the ones BI, AI, data science, and activation teams can actually build on.

Databricks & Unity Catalog administration

• Administer Unity Catalog: catalog bindings, external locations, storage credentials, grants, and the service-principal model behind them.


• Define workspace and compute standards — cluster policies, runtime versions, pool lifecycle, job compute versus all-purpose, and serverless where it pays for itself.


• Manage platform identity end to end: users, groups, SCIM provisioning from Entra ID, secret scopes, and token governance.


• Keep the estate current: runtime upgrades, deprecation tracking, and retiring the pools and jobs that quietly stop working.

Ingestion, orchestration & integration

• Design pipeline architecture across Azure Data Factory and Databricks Workflows: dependency design, idempotency, watermarking, and retry and backfill semantics that survive a bad source day.


• Keep source integrations healthy — Dataverse and Synapse Link exports, finance and booking systems, third-party feeds — including the runtime, connector, and credential lifecycle they silently depend on.


• Establish data contracts with upstream owners so schema changes reach the platform before they reach production.


• Land a modernization decision for SQL Server to Databricks without breaking the reporting built on it.

 AI-assisted engineering standards & harness enablement

• Establish and enforce the AI coding standard for the data team: what assistants may generate, what must be human-reviewed before merge, and what may never reach production without a test behind it.


• Build and maintain the AI coding harness — repository instruction files, project context, MCP servers, tool permissions, and sandboxed environments — so generated code arrives already matching our conventions.


• Establish review practice for generated code and SQL: correctness against the data model, cost and performance impact, and provenance recorded in the pull request.


• Continuously evaluate and improve output quality, refining instructions and context based on what the team actually gets wrong — treating the harness as a product with users, not a config file.

AI/ML platform administration

• Administer ML assets in Unity Catalog — registered models, aliases, feature tables, and MLflow experiments — including the grants and promotion path from development through QA to production.


• Define GPU and ML compute standards: node-type allowlists, Azure quota management, idle termination, spot strategy, and ML Runtime version lifecycle, so expensive compute cannot be left running unnoticed.


• Govern foundation model and external LLM access through the AI Gateway — rate limits, usage tracking, payload logging, guardrails, and provider credentials held in secret scopes rather than notebooks.


• Extend network and data-protection controls to AI workloads: egress rules for external model APIs, retention of logged prompts and responses, and handling of customer data in inference payloads.

Security, access & platform governance

• Define the platform's security posture: network isolation, private connectivity, storage firewalls, and a documented rationale for every public-access exception that remains.


• Run credential hygiene — secrets into Key Vault, scheduled rotation, and no long-lived plaintext tokens in notebooks, pipelines, or configuration.


• Design and administer the access model: role definitions, joiner/mover/leaver flow, privileged access, and periodic entitlement review.


• Work with the Infrastructure team and move platform configuration into Terraform and CI/CD and keep it there, so there is no drift between what is deployed and what is declared.

Identity resolution & marketing activation

• Contribute to a consent, suppression, and preference management solution, ensure rules are applied correctly and provably per market, and that a failure in that logic is caught by the platform rather than by a customer.


• Steward the data-sharing connections into Adobe Experience Platform and activation

Reliability, quality & cost

• Define SLAs for the pipelines the business actually depends on, then build the monitoring and alerting that proves you are meeting them.


• Lead incident response for platform failures: triage, root cause, remediation, and a written RCA that changes something.
  Close the class of failure that hurts most — silent partial loads, duplicate or missing records, and jobs that report success on incomplete data.


• Support the data quality framework: rule coverage, severity semantics, and making sure a blocking rule actually blocks.
  Monitor platform spend: cluster right-sizing, storage lifecycle, and a defensible view of cost per workload.