I build machine learning that stays reliable when deployment conditions violate training assumptions: sensor channels drop out, operating regimes shift, simulated signals diverge from measured ones, and the target hardware is far smaller than the trained model. I work through graph representation learning, physics-informed modelling, domain adaptation, and model compression, and I validate on operating systems rather than benchmarks alone.

• 🔧 Fault Diagnosis & PHM — graph convolutional models for rotating machinery under noise, class imbalance, missing data, and changing operating conditions
• 🚢 Maritime AI — a physics-constrained digital twin for vessel fuel consumption and route optimisation; controlled pilot across 10 vessels over 15 voyages cut fuel by 6.5% and CO₂ by 7%
• 🐄 Precision Livestock — behavioural anomaly detection from wearable accelerometers; per-animal dynamic baselines cut false alarms by 30% at comparable detection sensitivity
• 🌍 Non-Stationary Forecasting — corresponding author on a three-paper line applying deep learning to seismic time series (256 citations across the line)
• 🤖 Applied LLMs — a retrieval-augmented assistant deployed over internal legal and shipping-regulation documents, cutting routine enquiries to specialist departments by an estimated 30%
• 📦 Reproducibility — a public repository for every published paper, plus seven MIT-licensed data toolkits

Getting the model into production, not just into print.

Seven MIT-licensed Python packages that make the data underneath my work reproducible.

Machine health & seismic benchmarks — standardised acquisition, integrity checking, and leakage-aware splitting for the datasets behind all of my peer-reviewed publications.

Maritime & environmental data — reproducible geospatial and reanalysis extraction with checksums and source provenance.

Core ML/AI

Infrastructure & DevOps

Every paper below has a public repository; four have arXiv preprints.

In preparation (2026)

• Learned Objectives, Uncertain Routes: A Critical Review of the Coupling Between Data-Driven Fuel Models and Ship Route Optimization — in preparation for Ocean Engineering
• From Wearable Signals to Actionable Alarms: A Decision-Centric Review of Machine Learning and Optimisation for Behaviour, Calving and Health-Event Detection in Dairy Cattle — in preparation for Computers and Electronics in Agriculture

📚 Full list on Google Scholar

• 🎓 Ph.D. — Electrical Engineering (Control Systems), Tarbiat Modares University (2018–2023) · GPA 4.00/4.00, first-rank graduate
• 🎓 M.Sc. — Electrical Engineering (Control Systems), Tarbiat Modares University (2015–2018)
• 🎓 B.Sc. — Electrical Engineering (Telecommunications), Shahid Beheshti University (2010–2015)

Peer reviewer since 2022 — 70+ manuscripts for 11 journals, including IEEE Transactions on Industrial Informatics, Industrial Electronics, and Instrumentation and Measurement; Mechanical Systems and Signal Processing; Reliability Engineering & System Safety; Pattern Recognition; Neurocomputing; and Information Sciences.