Current themes
- LLM reasoning, evaluation, safety, and efficient inference
- Graph-based retrieval and multimodal memory
- Arabic NLP and linguistic evaluation
- Applied agentic systems for education, healthcare, drones, and scientific tooling
Latest highlights
2026 (5)
Because Flat Retrieval Is Not Enough: VideoGraph for Long Video Question Answering
Accepted at the ICML From Frames to Stories Workshop. VideoGraph improves long-video question answering with training-free persistent multimodal graph memory.
TAPS: Task Aware Proposal Distributions for Speculative Sampling
Accepted at the ICML SPIGM Workshop. TAPS studies draft-model training configurations for improving speculative decoding efficiency and generation quality.
AraGuard: Toward Safe and Reliable Arabic Large Language Models
Accepted at the ICML MusIML Workshop. AraGuard studies Arabic-first safety alignment and guardrails for large language models.
QuanBench Plus: A Unified Multi-Framework Benchmark for LLM-Based Quantum Code Generation
Accepted at ICLR ICBINB. QuanBench Plus evaluates LLM quantum code generation reliability across Qiskit, Pennylane, and Cirq.
AraLingBench: A Human-Annotated Benchmark for Evaluating Arabic Linguistic Capabilities of Large Language Models
Accepted at ACL/EACL. AraLingBench evaluates Arabic linguistic competence in LLMs using expert-curated benchmarks across core language categories.
Other published, accepted, and recent work
2026 (7)
Unified RGB-UAV Pipeline for Citrus Tree Detection, Geotagging, and HLB Health Assessment
Accepted at IGARSS. This work builds an end-to-end UAV pipeline for scalable citrus monitoring and disease assessment using RGB imagery and AI.
Toward Trustworthy Large Language Model Agents in Healthcare
Accepted at ICCBI. CareConnect uses safe LLM agents, RAG, and guardrails to automate healthcare scheduling with reliability constraints.
End-to-End Deep Learning Pipeline for Opportunistic Vertebral Fracture and Schmorl's Node Detection from Routine CT Scans
Accepted at Frontiers. This work develops an automated deep learning pipeline for detecting vertebral fractures and osteoporosis indicators from routine CT scans.
Real-Time Elevation and Orientation-Aware Visual Localization for GNSS-Denied Drone Navigation
Accepted at Drones by MDPI. REOLA supports GNSS-denied drone localization through visual matching and orientation estimation.
KAG: A Scalable Knowledge-Augmented Generation System for Educational Content Management
Published at FLLM. KAG combines vector databases and LLMs to improve educational retrieval and comprehension.
Chained Prompting for Better Systematic Review Search Strategies
Published at FLLM. This work uses a multi-stage chained prompting pipeline to generate transparent, high-recall systematic review search strategies.
PharmaBot: A Multimodal AI System for Drug Interaction Detection
Published at IMCET. PharmaBot detects and explains drug interactions using OCR, computer vision, LLMs, RAG, and microservices.
2025 (1)
Optimizing Deep Neural Networks using Safety-Guided Self Compression
Work on safety-guided compression methods for improving neural network efficiency under deployment constraints.
Work in progress
- Agentic Benchmark Project - active research project.
- Reasoning Vectors - targeting AAAI 2027; work in progress.
- UAV MARL - targeting AAAI 2027; work in progress.
- UAV Graph Plan - targeting AAAI 2027; work in progress.
- VideoGraph2 - targeting AAAI 2027; work in progress.
- Long Task CoT Benchmarks - targeting ICLR; work in progress.
- One-Shot CoT Reasoning - targeting ICLR; work in progress.
Earlier publications
2023 (1)
On Edge Human Action Recognition Using Radar-Based Sensing and Deep Learning
Published in IEEE Transactions on Industrial Informatics.
2022 (1)
A CNN-Based Method for Discriminating Shadowed Targets in FMCW Radar Systems
Published in Sensors.
2021 (1)
Maritime Localization System Based on IoT
Presented at the IEEE International Conference on Electronics, Circuits, and Systems.
2019 (1)
Convolutional Neural Networks for Arabic Font Recognition
Presented at the International Conference on Signal-Image Technology & Internet-Based Systems.