Case Study
Designing automated checking systems to support learner feedback, improve consistency, reduce repetitive review, and strengthen quality assurance across technical education.
Technical learners often make repeated mistakes across syntax, structure, naming, logic, and task requirements. Manually checking every submission can become time-consuming, especially when supporting multiple learners, cohorts, and assessment deadlines.
Feedback also needs to remain clear, fair, consistent, and aligned to the learning task. Without a structured review process, learners may receive uneven support or miss important next steps.
I developed automated code checking approaches that reviewed learner submissions against task requirements, common errors, expected keywords, structural patterns, and completion indicators.
The aim was not to replace instructor judgement, but to make the first stage of review faster, clearer, and more consistent, while keeping final feedback human-led.
Created checking logic that identified whether key task requirements were present in learner code, helping instructors quickly see what had been completed.
Built checks for recurring learner issues, including missing elements, incorrect syntax patterns, incomplete structures, and task-specific gaps.
Designed outputs that separated strengths, missing requirements, improvement points, and next steps so learners could act on the feedback more easily.
Used checking results to support evidence-led review, improve feedback consistency, and make assessment preparation more reliable.
Built an event-driven Google Apps Script workflow triggered automatically on learner submission, allowing reviews and feedback generation to happen immediately after form completion.
Developed structured checking logic for SQL requirements, syntax patterns, task completion indicators, extension activities, and common learner mistakes using reusable rule-based validation systems.
Added AI review detection logic combining hard evidence markers, soft phrasing analysis, score adjustment rules, instructor visibility controls, and structured AI outcome reporting.
Designed a queued email delivery system with retry handling, trigger automation, delivery tracking, idempotency protection, batch processing, and failure management to improve reliability across large learner groups.
Integrated automated RAG tracking, HTML feedback generation, spreadsheet validation, delivery state tracking, and assessment reporting to support scalable quality assurance workflows.
Structured the system to support instructor judgement rather than replace it, ensuring automated checks accelerated workflow efficiency while final assessment decisions remained human-led.
Repeated checking tasks became quicker to complete, allowing more time for targeted learner support and higher-value feedback.
Learner submissions could be reviewed against the same expectations, reducing variation across feedback and support.
Learners received more structured guidance on what was present, what was missing, and what needed to be improved.
The system supported clearer records, assessment alignment, and more reliable preparation for quality assurance review.
The next stage would be to develop this into a web-based learner review dashboard with secure uploads, version tracking, cohort analytics, reusable checking templates, instructor notes, and optional AI-assisted feedback drafting.