Projects
This page highlights some of the systems and studies the CORE Lab is building! We build scalable, adaptive systems that use artificial intelligence to personalize learning and the student experience. Each one starts as a question about how people learn in computing, engineering, and other computationally driven fields, from asking good questions and writing clearly to explaining their own reasoning and mastering concepts. That question then becomes a tool deployed in real courses and studied at scale.
Our systems share one approach: ground every interaction in the learning sciences, adapt to the individual learner rather than the average one, and instrument everything so that teaching and research improve together. If one could help in a course you teach, or you want to study learning on a platform already in classrooms, we would love to hear from you!
Sorted by papers, most first.
- QuickTACourse-aware chat tutoring with pedagogical guardrails and guided reflection, used by more than 15,000 post-secondary students across six disciplines and studied in randomized field experiments.11 people · 8 papers11 people8 papers
- VoiceExA voice-first reflection environment for studying how students explain their thinking, revisit difficult concepts, and respond to guided follow-up questions.7 people · 7 papers7 people7 papers
- Code to ConceptSynchronized code, memory, and analogy views for novice programmers, tested in a randomized trial with 829 CS1 students and extended to CS2 data structures.11 people · 6 papers11 people6 papers
- SQAMSQAM grades SQL by result set with partial credit, and weekly schema and data tasks with instant feedback (9,071 submissions from 669 students) build full data skills.11 people · 2 papers11 people2 papers
- NUMIA lab-built K–12 math practice platform studied with nearly 7,000 students in 20 schools: mastery requirements boosted practice, and the AI tutor improved recovery after errors.9 people · 1 paper9 people1 paper
- ARCAn integrated environment for systematic literature reviews, built around transparent iterative search and verifiable AI-assisted screening.7 people · 1 paper7 people1 paper
- MOSAICLLM-personalized learning materials built on the FACET framework. In CS1 they lifted completion above 99% and raised correctness 18% for the students most at risk.7 people · 1 paper7 people1 paper
- MarginA retrieval-augmented feedback pipeline for computing writing assignments. Students rated its feedback above TA feedback, and a cohort was processed in 1.8 hours instead of about 40.5 people · 1 paper5 people1 paper
- MedBotA voice-enabled simulated patient for medical learners, co-developed with clinicians, that scores each encounter against instructor rubrics and gives feedback.3 people · 1 paper3 people1 paper
- ReQuestIn developmentAn AI-supported trainer for requirements elicitation. Learners interview a simulated client, then receive feedback on question quality, coverage, and follow-up.5 people · 0 papers5 people0 papers
- CodeWalkthroughIn developmentA conversational code walkthrough for large courses: students explain their submitted code and answer follow-up questions until they meet the instructor’s learning objectives.4 people · 0 papers4 people0 papers
- Cul-SecIn developmentBuilding security into software from the first design decision, starting with SbDMLOps: security-by-design for trustworthy machine learning pipelines.3 people · 0 papers3 people0 papers
- Vis PipelineIn developmentCoding agents generate families of visualizations for programs, databases, and computer systems; each is verified against execution traces, and an adaptive policy chooses a design for each learner.3 people · 0 papers3 people0 papers
- CodeEvolverIn developmentA three-stage vulnerability review pipeline: a fine-tuned code classifier detects the flaw, retrieval grounds it in MITRE CWE guidance, and a small language model writes a report with line numbers and fixes.3 people · 0 papers3 people0 papers
- AMBERIn developmentAMBER blends feedback from multiple AI models, adapts it to each writer, and pairs every round of editing with guided reflection.2 people · 0 papers2 people0 papers
Other tools
Systems that CORE Lab members helped design, build, deploy, or evaluate, many in partnership with collaborators at other institutions. Each links to the paper that describes it.
Learning tools
- OnTrackReminder emails whose wording and timing are chosen by multi-armed bandits, tested in large courses.SIGITE 2022
- Pedagogical PromptingTeaches students to prompt LLMs in ways that support learning, tested in an RCT in CS1.arXiv 2026
- CodeAidA classroom LLM programming assistant that scaffolds rather than solves.CHI 2024
- CodetierlistA competitive, tier-list gamification platform for programming courses.ACE 2025
- ShapeCreator MobileAt-home coding practice designed for low-resource settings.ITiCSE 2025
- SSDVisInteractive visualization of operating-systems and SSD concepts.ITiCSE 2026
- AI-Generated TracesLLM-generated execution traces for novice programmers, studied across institutions.ITiCSE 2026
- On-Demand HintsHints and progress feedback on programming exercises, compared in large courses.ITiCSE 2025
- UML and ER Diagram FeedbackAutomated, criterion-based LLM feedback on student software-design diagrams.arXiv 2025
- Humanized Programming FeedbackGenerative models fine-tuned on student-written feedback to sound more like a person.arXiv 2025
- Adaptive Experiments PlatformInfrastructure for running adaptive experiments inside course platforms.LAK 2025
Research and writing tools
- ReflexisCollaborative qualitative analysis that supports researcher reflexivity.CHI 2026
- ScholarMateA mixed-initiative sensemaking tool for research literature.CHIWORK 2025
- TreeReaderAn LLM-powered hierarchical reader for research papers.VL/HCC 2025
- TreeWriterHierarchical AI planning and drafting for long documents.arXiv 2026
- ABScribeHuman–AI co-writing that lets writers explore many variations at once.CHI 2024
- CopilotLensA transparency layer that explains what AI coding agents did and why.arXiv 2025