Research Engineer I at MBZUAI · Specializing in LLMs, fact-checking & multimodal AI · 5+ years building AI solutions that matter.
I'm a Research Engineer I and NLP researcher at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) in Abu Dhabi, with 5+ years of hands-on experience in AI, machine learning, and NLP research. I hold an M.S. in NLP from MBZUAI (GPA 3.83/4.00).
My research centers on fact-checking and multimodal fact-checking — evaluating and improving the factual reliability of Large Language Models (LLMs) and Vision-Language Models (VLMs). I'm also passionate about temporal reasoning, spatial reasoning, and multilingual NLP, with a particular focus on Arabic and Urdu.
Beyond research, I bring extensive industry experience building scalable data pipelines, AI systems, and full-stack applications that have served enterprise clients across the US and UAE.
Mohamed bin Zayed University of AI (MBZUAI)
Supervisor: Prof. Preslav Nakov
GPA: 3.83 / 4.00University of Engineering & Technology (UET)
GPA: 3.49 / 4.00Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)
Singapore University of Technology and Design (SUTD)
GoPython Software Solutions → Prognos Health Inc
GoPython Software Solutions → Instant Labs Inc
Data Science Lab, Al-Khawarizmi Institute of Computer Science
Data Science Lab, KICS & Sabz Qalam
arXiv preprint · Annual Meeting of the ACL 2026.
CLEF 2026, LNCS 15087, pp. 465–492.
CLEF 2026 Working Notes — Jena, Germany.
The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) — Main Conference.
The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026).
The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026).
The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026).
The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP 2026).
The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) — Main Conference.
Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026).
European Conference on Information Retrieval (ECIR) 2026.
Findings of ACL: EMNLP 2025 — Suzhou, China.
Journal of Molecular Biology, 2025.
20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025).
Conference and Labs of the Evaluation Forum (CLEF) 2025.
Current Computer-Aided Drug Design, Vol. 20, Issue 6, pp. 773–783. DOI: 10.2174/1573409920666230817101913
Sabz Qalam, Vol. 1, p. 167. ISBN: 978-969-7941-00-1
Co-organized the 24th edition of ImageCLEF — the flagship multimodal evaluation lab at CLEF 2026. Covered 5 tracks: Medical (Caption, GANs, MEDIQA-CORE, MEDVQA), ToPicto, MultimodalReasoning, Deepfake Detection, and AI4Agriculture. Attracted 307 registered teams, 1,284 system runs, and 74 working notes papers — one of the most successful editions to date.
Reformulated medical fact-checking around snippet-level verification — grouping semantically related clauses into clinically coherent units rather than fragmenting into atoms. Built MedSNIP-Bench: 276 clinical-vignette responses, 2,524 human-annotated snippets with dual in-general and in-patient-context labels across 6 structural pattern codes. The MedSNIP pipeline reduces verifier LLM calls by 24–73% while preserving or improving F1 on HealthFC and MedHallu. Accepted at EMNLP 2026.
Curated a high-quality Urdu MCQ image dataset for multimodal reasoning. Developed baselines using VLMs and LLMs for cross-modal understanding in low-resource languages.
Agentic fact-checking framework for Urdu with evidence boosting, benchmarking, and LLM integration — accepted at EMNLP 2025.
Investigated limitations of ROUGE/entity metrics. Proposed structure-aware LLM evaluation at node, argument, and claim levels for long-form AI-generated content.
ML pipeline leveraging single-cell RNA sequencing data to classify COVID-19 severity. Used PCA + SVM; identified gene biomarkers (RSAD2, CXCL10, S100A8) via SHAP.
Evaluated hate speech detection in German, Hindi, and English using traditional ML and advanced LLMs, demonstrating LLM superiority in low-resource multilingual settings.
Ensemble of LLM embeddings and CNNs for peptide classification, achieving state-of-the-art results for Anti-Viral Peptides without handcrafted features.
Transfer learning with DenseNet-201 for binary classification of histopathology images, achieving ~87% accuracy across multiple magnification levels.
Evaluated 8 ML + 3 DL algorithms for multi-label emotion classification. Custom BERT with RoBERTa embeddings achieved macro F1 of 0.73 on Go-Emotions Dataset.
Cleaned and analyzed global COVID-19 data. Created interactive graphs and 3D geo-scatter plots to visualize global spread patterns over time.
React.js web app enabling users to visually create and connect complex database schemas. Paired with a Blockly-based tool generating SQL queries dynamically at runtime.
Fact-checking, temporal & spatial reasoning, multilingual benchmarking, agentic pipelines, and evaluation frameworks for LLMs.
End-to-end ML pipelines, OMOP CDM data modelling, AWS-driven automation, and enterprise healthcare data systems at scale.
Designing shared tasks, curating datasets, building evaluation pipelines, and publishing in top-tier NLP venues.
React/Go web apps, RESTful APIs, Docker/Kubernetes orchestration, and cloud-native solutions across AWS and Azure.
I'm always open to discussing new research opportunities, collaborations, or interesting projects in NLP and AI. Whether you have a question or just want to say hello, my inbox is always open.