AI Learning Engineer & Learning Experience Designer · learning science, supercharged with AI
I design learning experiences, and build the AI systems that let them scale.
Raised in Saipan. Trained as an educator and school counselor. Learned to build with AI at Amazon, shipping tools that replaced manual workflows. Now at EdPlus, directing Claude and Codex to build the learning tooling that lets the 5,000+ course ASU Online Course Catalog actually move.
“Learning is the pinnacle of human existence.”
the thing I keep coming back to
plate i.b.m.e.
Build the AI learning tools
Python + LLM pipelines that audit, generate, and remediate courses at the per-unit benchmark. Shipped at Amazon. Tested end-to-end at EdPlus.
Frame the problem before the model
Stakeholder discovery, scope, regression tests, post-launch metrics. AI is the tool; the workflow is the product.
Translate impact into evidence
Per-unit benchmarks, tested workflows, PartyRock analytics. Numbers tied to the system that produced them. No aggregate handwaving.
ID for the MS in Applied Statistics & Data Science. Hand-selected contributor to the Juris Doctor program team and a cross-functional AI taskforce (AI, UX, ID leadership).
Looking for
Learning teams that treat AI as core craft: building course-quality, authoring, and knowledge tooling, not just specifying it. Comfortable being the most technical person in a learning team, or the most learning-science-aware person in a technical one.
The receipts.
Some shipped, some validated at scale, some tested at the per-unit benchmark; aggregates labelled as estimates. Numbers tie to the system that produced them.
5h → 40s
per-course audit time
EdPlus Batch Course Evaluation pipeline: Python + Claude/Codex. Validated at scale on Canvas DEV against 1,000 real EdPlus course builds. Throughput 5,541 courses/hr (~133K/day). Zero failures across 71,000 deterministic evaluations. 100% data readiness using an adaptive rate-limiter.
Scope: building internal AI tooling for a 5,000+ course operation serving the entire ASU Online Course Catalog. Hand-selected contributor to the Juris Doctor program team and a cross-functional AI taskforce (AI, UX, ID leadership). ID for the MS in Applied Statistics & Data Science.
Batch Course Evaluation SystemPython + Claude + Codex on Railway; async workers, job queues, batching, retry logic
Per-course audit time 5 hours → 40 seconds. Validated at scale on Canvas DEV (asu-dev.instructure.com) against 1,000 real EdPlus course builds: 5,541 courses/hr (~133K/day) end-to-end throughput, zero failures across 71,000 deterministic evaluations, 100% data readiness on a 500-course representative run. Adaptive rate-limiter self-throttles against Canvas; cannot DoS the API. Grades against WCAG 2.1 AA + ID rubrics.
AI Course BuilderClaude-based authoring plugin, embedded WCAG + ID-standards checks, Canvas API integration
Total per-course build time, curriculum supplied: 110–140 person-hours / 3–4 weeks → 16–24 person-hours / 2–3 days. Per-course edit time 6 hours → 10 minutes. Validated end-to-end.
Audit-to-Remediation PipelineDetection, remediation queues, verification checks joined into one workflow
Closes the loop between detection and fix: audit findings flow into remediation queues with verification on the way out. Per-course review-fix-verify cycle reduced ~70% in tested workflow.
Shipped and ran in production for 5 months. Survey participation +56% (44.49% → 69.17%). 80+ hours/quarter/teammate of manual data collection eliminated. ~$50K/yr labor savings (estimated from time × loaded rate). Identified 20% of SOPs as out-of-date on first run.
AI-Integrated Learning SimulationsArticulate 360 with AI-driven branching
5/5 facilitator feedback. Trained 300+ associates with on-time delivery on all L&D submissions.
Amazon
Jul 2024 – Dec 2024
Logistics Specialist
6,500+ escalation cases at a 3% defect rate.
Mechanic On-The-Go
Dec 2022 – Apr 2024
Co-Owner
Operations and financial tracking for a retail automotive service.
CNMI Public School System
Jul 2023 – Apr 2024
School Counselor
Won a 20% budget increase via a data-driven grant proposal.
Things I’ve built.
Six systems and four live demos. Tools first; demos you can open below.
AI Tool
SOP-GPT
+30% performance on complex processes (PartyRock analytics).
AI Tool
LearnOps
+30% workflow efficiency for L&D (PartyRock analytics).
EdPlus
Batch Course Evaluation Pipeline
Per-course audit 5h → 40s. Validated at scale on Canvas DEV: 1,000 real EdPlus courses, 5,541/hr, zero failures across 71,000 evaluations.
Python + Slack webhooks + Asana + Amazon Quick Suite. Shipped and ran in production for 5 months. +56% survey participation; 80+ hours/quarter/teammate of manual collection eliminated; ~$50K/yr labor savings (estimated). 20% of SOPs flagged out-of-date on first run.
