Bio

I am an NLP & Generative AI researcher and Principal Applied Scientist at Oracle. My work focuses on LLM agent architectures, long-term memory, tool use, and evaluation, including how agents coordinate API calls and recover from failures. My broader research interests include aligning generative AI and building language technology for underrepresented languages.

Experience

Principal Applied ScientistAug 2025–present

Oracle

Research lead for long-term memory on OCI's Generative AI Agent Platform, in Dr. Yassine Benajiba's team. I design LLM agent architectures, tool-use capabilities, and evaluation infrastructure, and lead work from research through production delivery. Recent public work includes our long-term memory blog and our EMNLP 2026 Industry Track paper on enterprise cloud agents (co-first author).

Applied Research ScientistJan–Aug 2025

Grammarly

Designed and prototyped LLM systems for writing assistance and multi-agent orchestration. Led research on Agent-Testing Agent, an automated adversarial testing system for conversational agents (EACL 2026, oral). Sole supervisor of the research intern who co-authored the paper; coordinated analytical linguists and a dedicated annotation team.

Senior Research ScientistJul 2023–Jan 2025

TikTok · ByteDance Research

Led the Rephraser project for text-to-image prompt alignment, resulting in the Fast Prompt Alignment for Text-to-Image Generation technical report. Coordinated analytical linguists and a dedicated annotation team, set NLP research goals, published work on multimodal LLM evaluation, and co-supervised a PhD intern.

AI Research ScientistJun 2022–Jun 2023

Meta

Built production NLP and LLM systems for feed and hashtag recommendation, including work for Instagram and Facebook Reels, and for multimodal search. Delivered two product launches.

PhD internships

Four research internships across big-tech labs while at UC San Diego.

Adobe Research

Summer 2019

Developed a state-of-the-art parser and the Label Attention Layer.

Amazon Alexa

Summer 2020

Reinforcement learning for controllable abstractive summarization.

Meta AI

Summer 2021

Entity linking for news articles and user-written medical text, using multi-task learning for entailment, disambiguation, and entity recognition (then Facebook AI). See our ACL 2022 Findings paper.

Google Brain

Fall 2021

Task-oriented dialog systems with rewards (now Google DeepMind).

Education

Personal

I was born and raised in Marrakech, Morocco, a truly wonderful and lively city. I speak five languages; please feel free to reach out in any of them.

Explore my travel atlas: choose a country, or zoom in for cities, roads, lakes, and landscapes.