I build learning systems that scale.

The Problem

Five new learning programs needed to launch in the same window. Each one had to be built from scratch. No shared standards. No reusable architecture. Every team reinventing rubrics, design patterns, and review processes in parallel.

The operational drag was real. Build cycles dragged. Support tickets repeated themselves. Quality varied based on who built each course. At scale, this kind of duplication doesn't slow you down. It stops you.

What I Built

A learning design system with four operating layers.

A reusable template library covering blueprints, module patterns, assessment frames, and front-end components for visual consistency across the LMS. A QA system with pre-launch checklists, style guides, and compliance-ready documentation patterns. Standardized design patterns that mapped outcomes from the program level down to individual assessments, with accessibility built in by default. A repeatable review process with pre-launch audits, troubleshooting guides, and onboarding flows for new team members.

One architecture. Five programs running through it. Built once. Reused for years.

The Decision

I could have built five custom programs. Faster in the short term. Painful every time requirements shifted.

I built the architecture instead. One framework that all five programs ran through. Modality-agnostic. Compliance-ready. The decision cost more upfront. It saved everything downstream.

What I'd Do Differently

The system was solid. Adoption needed a stronger people layer. A formal champion model with one or two engaged team members per program would have accelerated rollout in the slower-moving programs.

I'd also build versioning in from day one. When requirements evolved, we had to retrofit changes across templates that weren't designed for easy iteration. Months of work that didn't need to happen.

The Problem

An organization needed to skill up 56 subject matter experts on AI tools, fast. No coherent learning path existed. No clear policy guardrails. No time to figure it out alone.

The result was predictable. Each expert quietly deciding for themselves what acceptable use looked like. Some experimenting. Some avoiding. None aligned. That works as a stopgap. It doesn't scale.

What I Built

A 9-module asynchronous learning path that experts can dip into based on what they need. Each module follows the same structure. Overview. Key concepts. Concrete examples. Optional deeper resources. The topics span AI-aware assessment design, AI for research and analytical work, ethical and responsible use, governance boundaries, and human-in-the-loop prompting frameworks.

The course is also a badge program. Subject matter experts who complete all 9 modules earn a credentialed badge they can display internally or on their LinkedIn — giving the course visible recognition and an incentive structure beyond completion alone.

The course isn't a high-stakes graded program. It's a resource hub. Learners open what they need. They skip what they don't. The course was the first of its kind in the organization.

The Decision

I could have built a comprehensive 12-week training. I built a self-paced resource hub instead. Experts don't have 12 weeks. They have 30 minutes between meetings. The hub structure respects that. The completion rates earned the design choice.

What I'd Do Differently

A 30-minute starter path from day one. A curated mini-sequence for experts completely new to AI. The modular hub works for people who already know what they need. Nine modules can paralyze someone starting fresh.

A living examples library populated by experts themselves. Real prompts. Real attempts. Real what-didn't-work. The course teaches principles. What people actually want is to see other people's real tries.

The Problem

End-users were already adopting AI tools faster than the organization could establish guidelines. Most without guidance. Some with quiet over-trust in the output. A few with active misuse.

Subject matter experts knew this was happening. They didn't have a consistent way to teach critical AI use. Some were restricting AI. Others were waving it through. Both approaches missed the actual issue. Future practitioners were forming AI judgment habits with no scaffolding.

What I Built

A standalone learning module that any program lead can copy into their workflow. Three layers. An introduction with learning objectives. A resources and strategies page covering tool access, practical prompting strategies (role, context, constraints), and what to watch out for (bias, hallucinations, overconfidence). And the heart of the module — a critical appraisal activity where learners compare AI-generated output against their own evidence-based reasoning.

End-users don't just learn to prompt. They learn to argue with the output.

The Decision

I could have built a quiz-based AI literacy module. I built a critical appraisal activity instead. A quiz tests recall. A critical appraisal task tests judgment. In high-stakes work, judgment is the only thing that matters.

What I'd Do Differently

A trust calibration exercise in week 1. Learners were assessing AI output well by week 2. Their baseline trust at week 1 was higher than I expected. A pre-test would have sharpened the learning arc and helped them notice their own assumptions sooner.

A companion resource for subject matter experts showing how the same critical-appraisal framework applies to their AI use. Workflow prep. Literature search. Writing feedback.

The Problem

Most learning programs treat assessment as the last step. Built once. Deployed forever. The reality is different. Assessment is the data layer of learning. Without high-quality outcome alignment, program leaders can't tell what's working, where learners are struggling, or what to fix.

In an organization launching new programs simultaneously, the absence of structured outcome mapping meant something more expensive. Every compliance cycle had to be rebuilt from scratch. Manual spreadsheets. Reconstructed evidence. Hours that nobody had.

What I Built

A curriculum mapping system that aligns every assessment to specific program and course outcomes, then surfaces the data program leaders need on demand. The system maps program outcomes down to course outcomes, course outcomes down to individual assignments, and assignments back up to assessment evidence. It supports two programs and 20+ courses.

The data was always there. The infrastructure made it usable.

The Decision

I could have built a one-time compliance report. I built reusable mapping infrastructure instead. Compliance cycles repeat. Manual reports require rebuilding. A mapped system makes compliance reporting a query, not a project.

What I'd Do Differently

Dashboarding from the start. The map exists. The data flows. What I'd add is a clean visual layer so program leaders can interpret the data without needing me as an interpreter.

A quarterly review rhythm with program leaders. The signal was always there. The collective ritual to act on it was the harder thing to build.

The Problem

Cultural awareness and inclusive practice were being taught unevenly across the organization. Some programs covered them deeply. Some touched them lightly. None shared a structured space for learners to engage with bias, cultural difference, and equity over time.

Senior leadership wanted more than memorized definitions. They wanted reflective practice as part of professional formation.

What I Built

A 4-module asynchronous learning experience that participants complete one module per cohort cycle, finishing all four by the end of their program. Each module covers a distinct theme — microaggressions and bias, disability and accessibility, socioeconomic factors and equity, and cultural competency. Inside each module are 6 to 7 content pieces and one or more reflection activities designed to provoke depth, not compliance.

The Decision

I could have built a single mandatory training. I built a 4-module longitudinal experience instead. A single training produces a checkbox. Longitudinal reflection produces practice.

What I'd Do Differently

Companion guides for program leaders. A 2-page companion per module would have multiplied impact without adding to my workload.