AI / Automation Engineer (Data Discovery & Classification)

Our client’s unstructured estate — thousands of SharePoint sites of varying ownership and activity, department drives, OneDrive, plus structured stores — cannot be mapped by interviews alone. This role stands up AI-enabled discovery and classification to find personal information, tag it to our client’s record types and populate the customer and associate data maps, with a human validation step at every stage. Phase 1 runs on a bounded dataset (and on our client’s existing Microsoft capability if external tooling has not yet cleared security review); Phase 2 scales across nine functions; Phases 3–4 hand over classification rules and re-scan to prove evidence traceability.

Nothing you build touches our client production customer or employee data until it has passed our client’s AI vendor/tool risk assessment and data processing and security review — you will help prepare that submission in week one. Output is never accepted on model confidence alone; the Data Engineer independently validates it.

What you will do, by phase

Deliverables you own or co-own

        Enterprise Data Inventory and Repository Register — AI discovery output presented alongside independently validated findings (co-owned with Data Engineer)

        Customer and Associate Data Maps — populated for every record type carrying personal information (co-owned with Compliance Business Analyst)

        Classification rule handover pack and re-scan evidence (Phases 3–4)

        AI tooling risk-assessment submission and approval record

Must-have qualifications

        5+ years applied machine learning, data science or automation engineering with hands-on data classification / sensitive-data discovery at enterprise scale

        Production experience with at least one discovery/classification platform: Microsoft Purview (sensitive info types, trainable classifiers, DSPM), BigID, Varonis, OneTrust or equivalent

        Designed human-in-the-loop validation with measurable quality (precision/recall, sampling plans, error analysis) and can explain results to non-technical reviewers

        Python; Microsoft Graph / SharePoint APIs or equivalent for estate-scale scanning; secure handling of results inside a client tenant

        Has prepared or contributed to an enterprise AI/security review (architecture, data flow, model/data residency) and shipped under its constraints

        Understands personal information categories under CCPA/CPRA, SPI and ePHI well enough to design classifiers for them

Strongly preferred

        Azure OpenAI / LLM-based document classification pipelines with evaluation harnesses; agentic workflows for governance documentation with review gates

        BigID or Purview certifications; NIST AI RMF familiarity

        Experience classifying legacy formats (mainframe extracts, COBOL copybooks, call-recording transcripts)

        Prior privacy or records program where classification output fed retention labels or a data map

Tools and platforms

Microsoft Purview Information Protection / DSPM, BigID, SharePoint Online and Microsoft Graph, Azure services as approved by the client, Python, Excel function workbooks, Clarity PPM.