What should an adaptive system infer?
Explore knowledge tracing and assessment while keeping prediction, mastery, validity, fairness, and learning impact conceptually distinct.
Research interests
Representing knowledge, time, and people responsibly.
I am developing a research program around educational technology and human-centered AI. The emphasis is on inspectable representations, appropriate evidence, and tools that help people reason about model behavior.
Scope note
This is an interests-and-artifacts page, not a publication record. The repository contains no verified publication list, so the items below are labeled as public prototypes, coursework, or technical explorations rather than peer-reviewed contributions.
Questions
What I am trying to understand
These themes connect the research repositories already represented in the portfolio with the educational tools maintained here.
How should changing state be represented?
Model interactions, concept relationships, user context, and time without collapsing meaningful structure into a static summary.
How can model behavior remain inspectable?
Build interfaces and explanations that expose assumptions, state transitions, tradeoffs, and failure conditions.
Public artifacts
Research and applied prototypes
Descriptions are intentionally bounded to the project intent recorded in this portfolio. Follow the repository links for implementation details and current status.
Temporal Heterogeneous Graph Knowledge Tracing
Explores temporal heterogeneous graph structures for student interactions and concept dependencies in knowledge-tracing workflows.
View repositoryHolistic User Map Starter
FastAPI reference implementation for persistent assistant personalization using compact, continuously updated user models.
View repositoryTemporal Subset ANN
Explores interval-valid temporal subset approximate-nearest-neighbor search with scalar and category filtering.
View repositoryBiomedical NER with CRAFT
Named-entity-recognition work using full-text biomedical articles and Gene Ontology annotations from the CRAFT corpus.
View repositoryClick-to-Talk
Voice-driven mouse interaction prototype intended to support people with fine-motor challenges.
View repositoryCustomer Data Pipeline
Data integration and analytics work spanning relational and NoSQL sources, ETL, warehouse modeling, and orchestration.
View repositoryResearch practice
Evidence before narrative.
Validity
Do not substitute accuracy for meaning.
Predictive performance alone does not establish mastery, fairness, validity, or improved learning.
Transparency
Expose the representation.
Make data, state, constraints, and decision paths visible enough to inspect and challenge.
Boundaries
Label prototypes honestly.
Separate demonstrated behavior from future work, unavailable infrastructure, and claims that have not been evaluated.