Diagnostic first, expository second, applied third
I treat teaching as part of the work, not an obligation beside it. Data science and AI now shape decisions in nearly every professional field, and careful instruction in these methods serves both the student in front of me and the public that will meet the systems they go on to build.
Three operating commitments
-
01
Artifact before definition
Every topic opens with something concrete: working code, a worked example, a small dataset. The principle is derived from what students have already watched happen, and a student-led extension is assigned in the same session.
-
02
Frequent low-stakes checks
Short retrieval problems open each meeting and are reviewed within five minutes. Misunderstandings surface before the first exam rather than after it.
-
03
Rubrics published up front
Solution rubrics go out with every assignment. Grade disputes collapse to a single exchange, and students know what the standard is while there is still time to meet it.
Active and experiential instruction
I use the flipped classroom model and think-pair-share alongside frequent in-class coding. Authentic research datasets from healthcare and climate informatics come into the curriculum, so students meet the structural and ethical complications of real data instead of the cleaned versions in textbook exercises.
Inclusive instructional design
Data science classrooms gather students at very different levels of preparation. I build multiple on-ramps: optional review modules, paired programming in lab, office hours framed explicitly as low-pressure space for foundational questions, and anonymous mid-semester surveys that arrive early enough to act on. For asynchronous content I record short modular videos rather than long single takes, so a student can revisit one technical segment without scrubbing through an hour.
Courses taught
Five consecutive semesters as a Graduate Teaching Assistant at UMBC across four undergraduate and graduate courses, January 2021 to May 2025, plus guest lectures in the UMBC Causal AI Lab from 2023 to 2026.
| Course | Terms | Responsibilities |
|---|---|---|
| Structured Systems Analysis and Design | Spring 2025 | Process modeling, design specifications, review sessions, grading |
| Advanced Database Project | Spring 2022 – Spring 2024 | Supervised graduate teams through requirements, schema design, indexing, query optimization |
| Database Program Development | Spring 2021; Spring 2022 – Spring 2024 | ER modeling, normalization, SQL, indexing, transactions; weekly recitations and labs |
| Management Information Systems | Spring 2022 – Spring 2024 | IT systems analysis, organizational decision making, individual and group tutoring |
| Guest lectures, UMBC Causal AI Lab | 2023 – 2026 | Causal discovery, time series modeling, trustworthy ML, delivered over Zoom |
Prepared to teach
Direct match, from TA experience.
Database Systems · Systems Analysis and Design · Software Engineering · Management Information Systems
Strong match, from training and research.
Introduction to Data Science · Introduction to Programming · Data Structures · Algorithms · Discrete Mathematics · Probability and Statistics for Computer Science
Graduate and advanced electives.
Artificial Intelligence · Machine Learning · Deep Learning · Data Mining · Time Series Analytics · Applied Machine Learning · Statistical Learning
Specialized electives I can develop.
Trustworthy AI · Causal AI for Decision Systems · Adversarial Machine Learning · Machine Learning for Intrusion Detection · Privacy-Preserving Machine Learning
Causal AI for Decision Systems
A complete graduate elective, built and ready for the first year of a faculty appointment. It covers structural causal models, do-calculus, time series causal discovery, counterfactual reasoning, and algorithmic fairness, with project tracks in healthcare informatics, climate science, and security analytics, and code released into the open-source ecosystem that supports my research.
The motivation is practical. Employers and collaborators increasingly need analysts who can recognize when a problem calls for causal reasoning rather than prediction, and few U.S. computer science or data science curricula treat these methods formally. The course differentiates a graduate program and hands students a publishable project pipeline.
Materials developed, a 44-slide module with embedded quizzes:
- Lecture transcript and technical notes on stationarity, AR and VAR models, Granger causality, and five open challenges in time series causal discovery
- Algorithmic treatment of four paradigms: tsFCI, NOTEARS and DYNOTEARS, TCDF, TTCD, and LCM
- Worked example applying TTCD to climate data
- Case study applying CDANs to ICU patient monitoring, with exercises built on the
cdanspackage - Curated reference list and in-class assessment items
What students leave with
In the UMBC Causal AI Lab I have co-authored peer-reviewed papers with other students at KDD, MLHC, AAAI, IEEE ICMLA, and IEEE PerCom, taking the lead on some and contributing a component to others. Alongside that I have worked informally with junior Ph.D. and M.S. students: scoping literature reviews, drafting experimental plans together, debugging pipelines line by line when results stall, and running mock conference talks before submission.
As faculty I will turn that into structured undergraduate research and capstone supervision: a weekly research clinic for active projects, resume and GitHub portfolio reviews keyed to the major, mock technical interviews twice a semester, and a graduate school application track for students considering an M.S. or Ph.D.
Open-source software for course use
The software from my research is designed for classroom use with no
licensing or implementation barrier: pip install cdans
for constraint-based discovery
(GitHub),
TimeGraph for benchmarking, and
Composable Causality (forthcoming) for end-to-end
method comparison. Course projects that contribute to these packages
produce work with currency outside the classroom.