Project: Ring-Vision
A field note about turning a repetitive home camera routine into an automated Ring camera capture and timelapse workflow.
Jason Mu He
Learning systems, data science, applied AI
Learning systems, data science, applied AI
I design enterprise learning systems, data pipelines, and production AI agents.
I work where instructional architecture, analytical data warehouses, and agentic AI meet. This site is a living record of what I am building, deploying, and evaluating in production.
Focus
Learning systems & AI engineering
Mode
Architect, build, deploy
Home base
Toronto, ON, Canada
From the Workshop
Project: GenAI SCORM Composer
A proof of concept for generating SCORM-compliant learning packages with LLMs, Streamlit, and Rustici SCORM Driver.
GenAI
SCORM
Authoring Tool
Rustici
SCORM Driver
A field note about using generative AI to speed up SCORM course creation while keeping the instructional design workflow visible.
See how it came togetherField Notes
What I Am Learning
Field notes on open questions, technical discoveries, and architectural decisions behind finished builds and experiments.
Project: GenAI SCORM Composer
A field note about using generative AI to speed up SCORM course creation while keeping the instructional design workflow visible.
GPT Search is Available for Everyone Now!
A short field note showing how to configure ChatGPT Search as a browser search shortcut.
Built Work
Things I Have Built
A curated trail of production data systems, enterprise automation engines, applied AI products, and research prototypes.
Agent PEB: Enterprise Data Operations Platform
An enterprise automated pipeline ingesting ERP files via SFTP, enforcing Pydantic row validation, reconciling state with Docebo LMS, ensuring transactional safety with Postgres advisory locks, and providing an AI Campus Copilot.
Python 3.12
PostgreSQL
Flask
Pydantic
Enterprise LMS
AI Copilot
AUTOCERT 2.0: Automated Certification Engine
A rules-based automation engine closing platform gaps and eliminating manual retraining tracking through daily cron workflows, large-scale enrollment data cleaning, and real-time compliance dashboards.
Automation
Rules Engine
LMS API
Data Quality
Python
GenAI SCORM Composer
An automated courseware workflow generating SCORM-compliant learning packages from learning goals using LLMs, Streamlit, and Rustici SCORM Driver to bridge instructional design and technical compliance.
GenAI
SCORM
EdTech
Python
Rustici Driver
Ontario Youth Badminton Analytics
A production sports analytics platform ingesting tournament match data, calculating rating trajectories, and serving access-controlled endpoints and public views with strict data-privacy boundaries.
FastAPI
Python
Supabase
PostgreSQL RLS
Next.js
Fuji Care / RUNJIA (润家·福吉)
A bilingual (English/Chinese) agentic wellness concierge and healthcare marketplace featuring native SwiftUI, Claude streaming with tool-calling events, VisionKit card scanning, and secure GCP Cloud Run backend.
SwiftUI
Claude API
HealthKit
Next.js
GCP Cloud Run
Post'Em Plus
A from-scratch Python/Flask rewrite of a legacy grading tool, providing secure, access-controlled student feedback, grade distribution, and LMS integration via standard LTI protocols.
LTI
LMS
Flask
Canvas
AWS S3
Credibility
Learning Systems & Data Engineering Practice
Over 15 years designing, architecting, and modernizing large-scale learning ecosystems, integrating cloud data warehouses, and building production-grade AI systems.
- Project Coordinator & LMS Site Administrator at the Centre for Addiction and Mental Health (CAMH, 2024–Present), engineering enterprise data pipelines (Agent PEB), rules engines (AutoCert 2.0), Snowflake-backed learning data warehouses, and clinical generative AI applications.
- Former Assistant Director in Learning Technology at Moody’s (2022–2024), leading enterprise learning modernization, 3-day cutover LMS migrations, and authoring published context-aware RAG research for self-directed study.
- Former Educational Technology Developer at the University of Delaware (2012–2022), architecting self-service Canvas automation integrated with ERP and AWS Redshift, and co-authoring LMS clickstream process-mining research.
- Doctor of Education (Ed.D.) in Educational Leadership & Technology, Master of Science (M.S.) in Data Science, and Master of Science (M.S.) in Instructional Design & Training.
Scholarly & Applied Research
Research & Publications
Peer-reviewed research and practitioner reports at the intersection of learning analytics, educational data mining, and context-aware generative AI.
The Path to AI SME: Harnessing Learning Analytics to Enlighten Conversational Agents for Self-Directed Study
He, M. • Companion Proceedings of the 14th International Conference on Learning Analytics & Knowledge (LAK24), pp. 21–24
Designed and evaluated a context-aware RAG architecture that conditions retrieval on learner LMS/LRS progress data before query generation. Achieved a 96.97% response rate at K=10 with ~49% lower token cost compared to conventional retrieval.
Retrieval-Augmented Generation (RAG)
Learning Analytics
Conversational AI
Vector Search
ACM / SoLAR
A Process Mining Framework for Analyzing Learning Clickstream Data
Wang, H., Wang, G., & He, M. • 10th China Summer Workshop on Information Management (CSWIM 2016), Dalian, China
Co-authored research proposing a process-mining framework for large-scale LMS clickstream data to model, discover, and analyze learner navigational sequences and behavior patterns.
Process Mining
Clickstream Analysis
Educational Data Mining
LMS Analytics
A Case Study of Designing and Developing Effective Educational App for Teaching and Learning Chinese as a Heritage Language
He, M. • Hawaii International Conference on Education (HICE 2015), Honolulu, HI
Applied case study examining the iterative design, implementation, and learning outcomes of a dedicated educational mobile app for heritage language acquisition.
EdTech Apps
Instructional Design
Mobile Learning
Now
Current Focus
I engineer robust enterprise data pipelines, applied agentic workflows, and high-reliability learning systems that bridge operational complexity and user experience.
Enterprise data operations pipelines & LLM-operable toolkits with safety guardrails
Context-aware RAG architectures & conversational agents for adult learning
Learning data warehouses (Snowflake, dimensional modeling) & real-time analytics
High-reliability LMS integrations (LTI 1.3, ERP delta reconciliation, advisory locks)