Education
PhD in Artificial Intelligence, University of Surrey
October 2022 - Present | EPSRC funded | Passed viva with minor corrections, September 2026
Thesis: Architecture and Objective Design for Robust Recommendation under Sparse Implicit Feedback. Supervisors: Prof Simon Hadfield, Dr Matthew Vowels, and Prof Liqun Chen.
MEng (Hons) Electronic Engineering with Nanotechnology, University of Surrey
September 2019 - June 2022 | First Class
Final project: Novel Design of Capacitive Tactile Sensor and Velostat Evaluation.
Electronic Engineering with Nanotechnology, University of York
September 2018 - June 2019 | Year 1, First Class (93%)
Research interests
- Recommender systems and sparse implicit feedback
- Multimodal and sequential learning
- Contrastive and graph representation learning
- Fair representation learning
- Optimisation, loss design, and distributional robustness
- Privacy-preserving machine learning
Research experience
PhD Researcher in Artificial Intelligence, University of Surrey
October 2022 - Present
- Lead EPSRC-funded research on architecture and objective design for robust machine learning under sparse implicit feedback, spanning sequential, multimodal and graph recommendation, neural ranking objectives, and fair representation learning.
- Build modular Python and PyTorch research codebases for data preparation, model implementation, training, evaluation, and analysis; run GPU-intensive experiments on University of Surrey HPC clusters using Slurm.
- Develop models that combine interaction sequences with collaborative, textual, and visual representations using transformers, graph neural networks, contrastive learning, attention, negative sampling, and sampled-softmax objectives.
- Own the end-to-end research workflow from literature review and problem formulation through implementation, error analysis, and publication; authored four first-author papers and reviewed for SIGIR, AAAI, ICML, ICLR, and RecSys from 2024 to 2026.
Professional experience
Software & Data Automation Engineer (part-time), Gulf Conferences Ltd
October 2022 - Present
- Design Python data-mining and automation workflows over more than 30 years of email history, structuring relationship data for targeted client re-engagement.
- Build web-scraping and data-processing pipelines to collect, clean, deduplicate, and structure prospective-client data for commercial outreach.
- Develop and maintain software, websites, and technical systems, automating recurring workflows and translating business requirements into working solutions.
Associate Software Engineer, Gulf Conferences Ltd
June 2019 - September 2022
- Developed and maintained websites, digital tools, and software workflows supporting business and event operations.
- Built automation scripts for recurring data-processing tasks and troubleshot software and live-system issues.
Selected projects
Technical skills
Programming and ML: Python, PyTorch, TensorFlow, C/C++, MATLAB, Pandas,
scikit-learn
ML and modelling: Transformers, graph neural networks, contrastive learning,
representation learning, sequential and multimodal recommendation, collaborative filtering,
neural ranking, sampled softmax, and negative sampling
Research engineering and compute: Slurm, GPU/HPC clusters, data
preprocessing, reproducible training and evaluation pipelines, hyperparameter tuning,
baseline reproduction, A/B testing, ablation studies, and robustness and distribution-shift
analysis
Evaluation and communication: Recall@K, NDCG@K, fairness evaluation,
statistical analysis, experimental design, technical writing and presentations, peer review,
and stakeholder communication
Selected publications and manuscripts
- Differential Adjusted Parity for Learning Fair Representations. Accepted at the ICLR AFAA Workshop, 2026. arXiv / PDF
- DSL: Understanding and Improving Softmax Recommender Systems with Competition-Aware Scaling. Under review; arXiv:2602.07206, 2026. arXiv / PDF
- Multimodal Enhancement of Sequential Recommendation. Under review; arXiv:2602.07207, 2026. arXiv / PDF
- Sequences as Nodes for Contrastive Multimodal Graph Recommendation. Under review; arXiv:2602.07208, 2026. arXiv / PDF
Awards and funding
EPSRC-funded doctoral research, University of Surrey.
Languages
English and Arabic.
Contact
London, UK
Email: bsahiony@gmail.com