I build andevaluateAI systems.

My research covers language models, legal reasoning, model security, and public-interest applications. I study how these systems work, where they fail, and how they can be improved.

Explore the research afane@law.stanford.edu

01 / PALO ALTO

Data scientist & AI researcher at Stanford RegLab.Working across machine learning, law, and public policy.

01 · Selected research

Selected work across reasoning, retrieval, and reliable AI.

ICAIL 2026Peter Jackson Best Innovative Application Paper

Learning When Not to Decide

A framework for overcoming factual presumptuousness in AI adjudication

Leading AI systems issued decisions in cases missing critical legal facts. We introduced SPEC, a structured method that identifies missing evidence before allowing a determination.

Read the paper with Emily Robitschek, Derek Ouyang & Daniel E. Ho

Determination–deferral tradeoffPaper · Figure 6

89 / 89

Accuracy on complete / inconclusive cases

ACM CS&Law 2026Legal RAG

Benchmarking
Legal RAG

The promise and limits of AI statutory surveys

We evaluated legal research tools across all 50 states. STARA outperformed commercial systems and surfaced valid laws missing from the federal expert compilation used as ground truth.

Read the paper Presented at UC Berkeley, March 2026

Statutory survey accuracy50 states

5060708090100%

Westlaw AI58

Lexis+ AI64

Standard RAG70

STARA90%

90% accuracy across the 50-state statutory survey benchmark.

Journal of Urban Health · 2025Public health

Lead Testing
in New York City

Finding a fairer distribution of childhood blood lead testing

We analyzed testing and elevated blood lead levels across 42 New York City neighborhoods, then optimized resource allocation toward underserved areas with greater need.

Read the paper with Juntao Chen

Map of New York City neighborhoods grouped into average, low, declining, rising, and high blood lead level risk profiles. — **14.3%**projected increase in cases detected through optimized testing allocationPublished figure: final neighborhood clustering

02 · Publications

Across legal AI, trustworthy machine learning, security, and public health.

Google Scholar

2026Legal AILearning When Not to DecideICAIL · Peter Jackson Best Innovative Application Paper 2026Legal RAGBenchmarking Legal RAGACM Symposium on Computer Science and Law 2025AI SecuritySCOUT: A Defense Against Data PoisoningIEEE International Conference on Big Data 2025Public HealthCan LLMs Help Allocate Public Health Resources?A case study on childhood lead testing · IEEE Big Data 2025Public HealthOptimizing Blood Lead Level Testing in New York CityJournal of Urban Health 2024AI SecurityNext-Generation PhishingIEEE International Conference on Big Data 2024Machine LearningSelecting Classifiers and Resampling Techniques28th International KES Conference

03 · Recognition

Selected honors.

2026

Peter Jackson Best Innovative Application Paper Award

International Conference on Artificial Intelligence and Law

2024

Third Place

Fordham Three Minute Thesis Competition

2023

GSAS Centennial Scholarship

Fordham University

2021—23

Fulbright Scholarship

U.S. Department of State

2015—20

National Scholarship

Mauritania

04 · About

Research across models, methods, and public systems.

I am a data scientist and AI researcher at Stanford RegLab, where I build and evaluate machine learning systems for complex, consequential settings.

My work includes legal AI, model evaluation, security, and public health. Across these domains, I am interested in systems that can be tested rigorously and used responsibly.

Education

M.S., Data Science · Fordham University

M.S., Energy Engineering · University of Illinois Chicago · Fulbright

B.Eng., Electrical Engineering · ENSAM Casablanca