Multi-agent LLM framework predicts one-year risk across multiple cancers
TrajOnco estimates one-year risk across 15 cancer types directly from longitudinal EHR, with AUROCs of 0.64–0.80.
2026 · LLM · Oncology · ASCO 2026
Writing
Research write-ups and project reports.
Research Write-ups
Zero-shot chain-of-agents predicts one-year lung cancer risk from longitudinal EHR
AUROC 0.871 with no model training and no feature engineering, while remaining clinically interpretable.
2026 · LLM · Oncology · ISPOR 2026
Patient journey foundation models improve EHR data quality
98% accuracy imputing missing or incorrect units of measure across 438 unique UCUM units.
2026 · Foundation models · Data quality · ISPOR 2026
Characterizing oncology patient journeys and health state transitions using a data-driven Markov transition matrix in large-scale electronic health records
Deriving oncology patient-journey state transitions empirically from longitudinal records.
2026 · Markov models · Oncology · ISPOR 2026
Can a generative patient journey foundation model alleviate the burden of cancer screening?
A foundation model evaluated on 1.42M patient records reached 0.89–0.98 specificity across four cancer types with no labeled training data.
2026 · Foundation models · Screening · ISPOR 2026
Predicting early-onset colorectal cancer with large language models
The AMIA proceedings work: a fine-tuned LLM reached 73% sensitivity at 91% specificity in adults aged 18–44.
2025 · LLM · EHR · AMIA 2025
Machine learning risk prediction for early-onset colorectal cancer
Development and validation across 30 US health systems, presented at ASCO 2025.
2025 · ML · EHR · ASCO 2025
Fine-tuning ChatGPT for essay grading
Third in a series on automated essay scoring: what fine-tuning buys you over prompting.
2024 · LLM · Assessment
The art of crafting prompts for essay grading
Prompt design for automated essay scoring, and how much of the variance it actually explains.
2024 · LLM · Assessment
A comprehensive guide to essay grading with ChatGPT
End-to-end walkthrough of grading and classifying essays via the ChatGPT API, with cost and runtime comparisons against human raters.
2024 · LLM · Assessment
Earlier Coursework
ENSO prediction with deep learning
Comparing deep learning architectures for forecasting the El Niño–Southern Oscillation.
2022 · Deep learning · Time series
Educational process mining
A web tool that turns raw learning-platform log data into something students and instructors can act on.
2021 · Learning analytics · Web tool
Perceived academic ability and adolescent depression
Time-series modeling of the Korean Children and Youth Panel Survey, 2010–2016.
2020 · Longitudinal · Psychometrics
Trajectories of socioeconomic status, self-control, and school readiness
Structural equation modeling of parent–child interaction and academic performance. AERA conference poster.
2019 · SEM · Longitudinal
Caregiver aggravation, child relationships, and academic performance
Analysis of the Fragile Families and Child Wellbeing Study.
2019 · Python · Survey data
Chronic absenteeism in rural India
Multilevel modeling of biometric daily-attendance data to identify socioeconomic predictors of absenteeism.
2019 · Multilevel modeling
No matching items