Reference Projects
No client names. No logos. The work speaks for itself.
Every project below was built and deployed in an active manufacturing environment — not piloted, not prototyped, not handed off to IT. Operational from day one.
AI-Driven Production Scheduling Engine
The problem: A multi-line manufacturing environment running production scheduling through a combination of ERP outputs, manual spreadsheets, and tribal knowledge. Schedule accuracy degraded daily as variables compounded — material availability, labor, machine state, work order priority — and no single source of truth existed for what was actually running versus what was planned.
What was built: An AI-driven scheduling engine integrating real-time shop floor data with ERP outputs (JDE/AS400 via ODBC) across multiple production lines. The system runs on Microsoft Fabric and Azure, processes live operational data via PySpark, and surfaces shift-level schedule accuracy through Power BI. Manual scheduling intervention dropped to exception-handling only.
Stack: Microsoft Fabric · PySpark · Azure · Power BI · Power Apps · SharePoint · JDE/AS400 ODBC
Status: Deployed in production. Active development continues on edge-case handling and multi-site expansion.
3-2-1 Training Framework
The problem: A manufacturing training program built on tribal knowledge, informal OJT, and documentation that varied by who wrote it and when. Qualification standards were inconsistent across shifts and lines. There was no mechanism to verify that a qualified operator was actually qualified — only that someone said so.
What was built: A standardized four-section training architecture applied across approximately 100 role-specific documents: On-the-Job Training, Qualification, Certification, and ReCertification. Each document follows the same structure, uses the same verifiable performance standards, and produces the same output regardless of who delivers the training. The framework is maintainable by internal staff without outside support.
Scale: ~100 documents rebuilt or created from scratch. Deployed across multiple production sites and job families.
Result: Consistent qualification standards, auditable training records, and a clear progression path for every trained role.
Workforce Intelligence System
The problem: Workforce data existed — in Workday, in spreadsheets, in SharePoint lists, in people’s heads — but no one could answer operational questions from it. Who’s qualified on Line 4 nights? What’s the bench depth on the press room? If attrition holds at current rates, what does Q3 headcount look like by line?
What was built: A workforce intelligence infrastructure connecting HR records, training qualification status, and operational assignments into a single queryable data model. Power BI dashboards built for shift leads and operations managers — not HR — surface skill gaps, bench depth by role, and attrition trajectory in operational terms.
Stack: Power BI · SharePoint · Workday integration · custom data model
Result: Workforce decisions made on data instead of assumptions. Critical role gaps visible weeks before they become emergency backfills.
Critical Role Continuity Program
The problem: A standard succession plan existed on paper. It didn’t reflect actual qualification status, hadn’t been updated in over a year, and identified “successors” who had never been assessed against the role requirements they were supposed to fill.
What was built: A structured continuity program identifying critical roles by operational impact, assessing actual bench depth against defined competency requirements, and producing a risk-ranked readiness register with remediation sequencing. Not a replacement for HR succession planning — a parallel operational layer that answers the question: if this person is gone Monday, what breaks and how fast?
Result: Critical roles ranked by coverage risk. Remediation priorities sequenced by operational exposure, not seniority or preference. Gaps addressed before they became failures.
Have a problem that looks like one of these?
Engagements start with a scoping conversation — no proposal before the call.