Ephraim Teodoro

Problem

The research needed more than prompt experiments: it needed a working system that could process essays, preserve evidence, support human decision notes, and export completed workbook results.

Solution

I built the technical system around an Electron desktop frontend, a local HTTP API, and a Python backend that coordinates NLP tools, local LLM inference, scoring progress, review screens, and workbook export.

My Role

Primary technical system builder for the research framework, translating the study requirements into a local/offline application workflow.

Result

Implemented and ready for experimental use. Associated research accepted for presentation at the 23rd AsiaCALL International Conference in Hanoi, Vietnam, November 14-15, 2026.

Problem

Learners need targeted practice that responds to weak characters, due reviews, recent mistakes, and confusion pairs without relying on a one-size-fits-all lesson order.

Solution

I built a Next.js 14 and TypeScript application with localStorage persistence for anonymous users, MongoDB persistence for authenticated users, and Google federated login.

My Role

Full-stack developer for the learning interface, persistence model, learner analytics, and rule-based recommendation workflow.

Result

Implemented as a rule-based adaptive learning system with deterministic recommendations, not machine learning or generative AI.

Problem

TeamDynamix does not have a native club concept, so retrieving tickets for a club previously required three external TeamDynamix API requests.

Solution

I designed a background synchronization service that stores TeamDynamix ticket data in MySQL, allowing the application to retrieve a club's tickets through one database-backed request.

My Role

Lead developer and project manager, coordinating development with the team and Student Leadership Services while contributing backend APIs, integrations, and database workflows.

Result

3 external TeamDynamix API requests -> 1 database-backed request, a 67% reduction in external API calls.

Availability

Internal workplace application with proprietary code and no public demo or repository. For department inquiries, visitors can contact BYUH Enterprise Information Systems.

Problem

The study investigated how standard and bubble scatter plots affected pattern recognition, population comparison, and outlier detection using United Nations tourism data.

Solution

I combined D3.js visualization work, Neon PostgreSQL-backed data handling, experimental design, and statistical analysis in a 30-participant counterbalanced within-subjects study.

My Role

Research software contributor working across visualization implementation, experiment design, data handling, and analysis.

Result

Found that bubble scatter plots improved pattern recognition and population comparison, while standard scatter plots performed better for outlier detection.