I'm an AI Research Scientist based in Paris, working on process-level evaluation of AI systems: LLM reasoning, multimodal perception, and retrieval quality. I hold a PhD from the Université d'Orléans (2024) and have published 20+ papers at venues including CVPR, ETRA, ICIP, and IEEE Access.

Experience

AI Research Scientist & AI Ops Architect

Jan 2025 to Present

F-Initiatives / NeoPhi · Paris, France

  • Shipped the production citation-verification pipeline for a scientific-literature platform indexing 100M+ papers
  • Raised RAG retrieval F1 from 0.13 to 0.50, a 4–6× relevance gain on private benchmarks
  • Distilled Qwen3.5-0.8B via merge-corrected iterative SFT (+3.7 pts GSM8K over base, MMLU retained); released on the Hugging Face Hub
  • Built year-aware venue ranking (SCImago × CORE), attributing Q-ranks to 100M+ papers

Technical Lead, HeReFaNMi Project

May 2023 to Nov 2024

F-Initiatives / IULM University Milan · Orléans, France

  • Secured EUR 130,000 in EU research funding (NGI Search, European Commission)
  • Architected and shipped a production health-misinformation detection platform
  • Led a cross-functional team of 4

Doctoral Researcher

Jan 2020 to Feb 2024

Laboratoire PRISME, Université d'Orléans · Orléans, France

  • 15+ peer-reviewed papers (CVPR, ETRA, ICIP, IEEE Access) on visual attention and scanpath modelling
  • Ran eye-tracking studies with 150+ participants across 10 museums
  • Secured EUR 100,000 across two projects (Louvre; INSERM)
  • Supervised 7 research interns

Teaching Associate (ATER)

Sep 2023 to Sep 2024

Polytech Orléans · Orléans, France

Taught and supervised lab sessions in computer science and machine learning.

Education

Ph.D., Computer Science & Artificial Intelligence

2020 to 2024

Université d'Orléans, France

Thesis: Visual Attention Prediction and Scanpath Modelling for Paintings.

Postgraduate Diploma, Data Science & Data Engineering

2022

Université d'Orléans, France

Oxford Machine Learning Summer School (OxML)

2024

Representation Learning & Generative AI

Selected Open-Source Projects

Iterative SFT distillation with automated model-merge selection across seven strategies.

APEX

GitHub

Agentic image-editing system with MLLM orchestration and identity-preservation gating.

Production health-misinformation detection: microservices, hybrid RAG, and vLLM.

FDISS

ETRA 2026 LBW

Perceptually grounded scanpath-similarity metric (Foveal Disc IoU).

SSLArtScanpath

CVPR-W 2022

Barlow Twins self-supervised scanpath prediction on paintings.

Browse all projects

Selected Publications

Lead author · under review (AAAI 2027 in prep.)

What They Saw, Not Just Where They Looked: Semantic Scanpath Similarity via VLMs and NLP Metrics

ETRA 2026 W

Lead author

Closing the Foveal Gap: Perceptually Grounded Scanpath Comparison with Disc IoU

ETRA 2026 LBW

Lead author

Self-Supervised Scanpath Prediction Framework for Painting Images

CVPR-W 2022

Lead author

All publications Google Scholar

Impact

197Citations
8h-index
7i10-index
23+Publications

Skills

ML & Training

PyTorch · Hugging Face (Transformers, TRL, PEFT, Accelerate) · vLLM · Model Merging · MLflow · Weights & Biases

NLP & Retrieval

Hybrid retrieval (BM25s, RRF) · Rerankers · NLI · LangChain · LlamaIndex · ChromaDB

Infrastructure

Python · FastAPI · Docker · SQL · React · TypeScript

Research Methods

Experimental design · statistical analysis (mixed-effects, signal detection theory, bootstrap CIs)

Languages

Arabic (native) · French (fluent) · English (fluent)