American University of Beirut · MSFEA
Serious AI research.
A place to begin.
AUB Frontier AI Fellows brings undergraduates into small research teams, with AUB faculty supervision and Lebanese AI mentors from frontier labs, universities, and companies.
Earn academic credit. Learn the craft of research. Contribute to questions that matter.
01 / The program
Learn research
by doing it.
A credit-bearing research program within MSFEA’s Vertically Integrated Projects Program (VIP).
Build the habits behind serious AI research: read papers, reproduce results, implement ideas, run experiments, and communicate what you find. Work toward publication with sustained effort and structured supervision.
Close mentorship
A small team.
A shared research question.
- 1
- AUB faculty supervisor
- 1
- Lebanese AI mentorfrom a frontier lab, university, or company
- 1–2
- AUB undergraduate students
The faculty supervisor owns academic oversight. The Lebanese AI mentor supports research direction and external perspective. Small groups keep supervision personal and manageable.
Your path through the program
Join a research course
Contribute to a defined AI research effort through a VIP course.
Present your work
At the end of each cohort, share what you worked on, what you learned, and the evidence of progress.
Earn the next step
Continue after review of your participation, output quality, reliability, and fit for the next phase.
Up to 6 credits in total. Progression to each course follows presentation and review.
02 / Previous work · 2026
A showcase of
previous research.
Previously published or accepted work coauthored by AUB faculty and collaborators, spanning efficient inference, quantum computing, Arabic language evaluation, and poetry generation.
Explore each paper for its findings, collaborators, and source.
TAPS Efficient language-model inference
TAPS: Task Aware Proposal Distributions for Speculative Sampling
An AUB–KAUST collaboration studying how task-specific draft models can make language-model generation more efficient, and how training-data choices affect the speed and quality of speculative decoding.
Collaborators: Mohammad Zbeeb, Mohamad Bazzi, Hasan Abed Al Kader Hammoud, Bernard Ghanem, and AUB faculty.
Read the accepted SPIGM workshop paperQuanBench Plus Evaluating quantum code generation
QuanBench Plus: A Unified Multi-Framework Benchmark for LLM-Based Quantum Code Generation
An AUB–KAUST benchmark testing language-model-generated quantum programs across Qiskit, PennyLane, and Cirq, using 42 aligned tasks and executable correctness checks.
Collaborators: Ali Slim, Haydar Hamieh, Jawad Kotaich, Yehya Ghosn, Mahdi Chehimi, Hasan Abed Al Kader Hammoud, Bernard Ghanem, and AUB faculty.
Read the accepted ICBINB workshop paperAraLingBench Measuring Arabic language capabilities
AraLingBench: A Human-Annotated Benchmark for Evaluating Arabic Linguistic Capabilities of Large Language Models
A benchmark of 150 expert-written questions assessing Arabic grammar, morphology, spelling, reading comprehension, and syntax, evaluated across 35 language models. Published in the AbjadNLP workshop proceedings in ACL Anthology.
Collaborators: Mohamad Zbib, Hasan Abed Al Kader Hammoud, Sina Mukalled, Nadine Rizk, Fatima Karnib, Issam Lakkis, Bernard Ghanem, and AUB faculty.
Read AraLingBench in ACL AnthologyShaer Controlled Arabic poetry generation
Shaer: Controlled Arabic Poetry Generation with Meter Subform and Semantic Conditioning
A system for generating Classical Arabic poetry with control over poetic meter, meter subform, semantic description, and poem length. Accepted to ArabicNLP 2026, co-located with EMNLP.
Collaborators: Ahmad Abbas, Tamara Fakih, Nour Fakih, and AUB faculty.
View Shaer on the ArabicNLP accepted-paper list03 / Who it’s for
Consistency comes
before expertise.
For current AUB undergraduates ready to commit to research.
Students in CCE, CSE, ECE, ME, Computer Science, and related fields do not need a publication record to start. Selection will be very competitive; strong motivation and consistent effort matter.
- 01Curiosity about modern AI research questions
- 02Comfort learning new technical tools quickly
- 03Clear written updates and follow-through
- 04Willingness to read papers, reproduce results, and ask precise questions
The next cohort
Bring your curiosity.
Make a start.
The first credit-bearing VIP cohort is expected to start in Spring 2026–27. Interviews and shortlisting begin in November 2026.
Submit an expression of interestQuestions? Contact AUB faculty at am288@aub.edu.lb.