CV

Posted on Mar 26, 2026

bayanabusalameh@gmail.com · linkedin.com/in/bayanabusalameh · github.com/Bees996

Education

Harvard University — Cambridge, USA (Aug. 2025) NSF-IAIFI Summer School and Workshop

National University of Singapore — Singapore (Aug. 2024 – Present) PhD, Mechanical Engineering. Thesis: Identification of Nonlinear Dynamics using ML.

Imperial College London — London, UK (May 2022 – Sep. 2024) PhD (research transfer), Mechanical Engineering.

Queen Mary University of London — London, UK (Sep. 2020 – Aug. 2021) MSc Advanced Mechanical Engineering, Distinction (CGPA: 4.0/4.0). Thesis: 1U CubeSat Design.

Birzeit University — Birzeit, Palestine (Oct. 2016 – Sep. 2019) BSc Mechanical Engineering, (CGPA 3.3/4.0). Thesis: Modified Tire-shredding machine.

Experience

Imperial College London — Graduate Teaching Assistant, London, UK (Oct. 2022 – Aug. 2023)

  • Taught and facilitated seminars, tutorials, and labs for ME2 Thermodynamics, Math and Computing, Mechanics, and Vibrations and Dynamics (20–30 students), integrating theory with hands-on experiments, problem-solving, and numerical computation.
  • Supervised and mentored undergraduate and master’s students on projects (computational methods, PDEs, vibration analysis), provided assessment and constructive feedback, and managed lab demonstrations.

NVIDIA, Inc — Mechanical Engineer, Yokneam (Aug. 2019 – Nov. 2020)

  • Led design and validation of 400 Gbps transceiver test boards (OSFP specification) and thermal management solutions; developed automated data-analysis scripts (Python) and test automation for high-throughput validation.
  • Developed and optimised mechanical and electro-mechanical packaging using 3D modelling (Inventor) and advanced manufacturing (CNC, injection moulding), collaborating with cross-functional teams and vendors.

RoboticX — Consultant Engineer, Birzeit, Palestine (Oct. 2018 – Jul. 2019)

  • Consulted on product development projects; specified budgets, assembly procedures, and introduced an in-house IoT automation system to improve operational efficiency.

Selected Projects

SDOF Nonlinearities Benchmark & Dataset Julia · MATLAB · Python · PyTorch · JAX

  • Constructed nine labelled synthetic datasets of SDOF displacement time-series (LNM1, LNJ1, DM1, DJ1, DJ19, ONM19, OWM19, ONJ19, OWJ19) capturing cubic stiffness (hardening/softening), Coulomb friction, clearance, and quadratic damping using Julia and MATLAB.
  • Trained and compared deep-learning classifiers (TCN, BiLSTM, CNN) directly on raw displacement signals — no domain-specific preprocessing — achieving classification accuracies of up to 99.8%.
  • Conducted post-hoc interpretability analysis (Integrated Gradients, GradientSHAP, SHAP) showing strong correspondence between model-important signal regions and expert-recognised features such as frequency-response distortions; framework targets structural health monitoring and dynamic characterisation of mission-critical systems including CubeSats.

Explainable AI for Structural Nonlinearity Classification Python · PyTorch · Captum · SHAP

  • Developed interpretability workflows (Integrated Gradients, GradientSHAP, DeepLift/DeepLiftSHAP, and SHAP) to map learnt features to physically meaningful signatures in time-domain responses.
  • Produced saliency/relevance maps and FRF-based interpretability visualisations that link network decisions to modal and energetic distortions, enabling engineering validation and improved trust in model outputs.
  • Packaged interpretability routines as reusable modules and used explanations to guide dataset augmentation and improved classifier generalisation.

ECG Out-of-Distribution (OOD) & Outlier Benchmark Python · PyTorch · NumPy

  • Created an end-to-end benchmark for OOD and outlier detection on ECG-like signals; managed raw datasets stored as signals.npy / targets.npy and designed preprocessing, augmentation, and OOD generation strategies.
  • Evaluated multiple detection methods (EWMA/thresholding, ensemble uncertainty, density-estimation, Mahalanobis/feature-distance, and deep-metric approaches) using AUROC, FPR@95%TPR, and precision/recall tradeoffs.
  • Delivered reproducible training and evaluation pipelines with logged experiment artifacts, example notebooks, and clear replication instructions.

Nonlinear Mode Interaction (NMI) Benchmark Python · MATLAB · PyTorch

  • Introduced a physics-based benchmark of 43,200 labelled multichannel time-series samples (displacement, velocity, acceleration) derived from 14,400 clean modal simulations of 960 unique 2-DOF nonlinear oscillators.
  • Covered detuning configurations near 1:1, 2:1, and 3:1 internal resonance ratios across multiple damping regimes, nonlinearity tiers, and excitation protocols; interaction labels (Linear/Non-interacting, Weak, Strong, Deep) assigned via a clean-label physics-based diagnostic pipeline in modal coordinates.
  • Corrupted clean simulations with controlled additive noise (25, 20, and 15 dB SNR) to emulate realistic experimental conditions, supporting systematic ML model comparison for robustness, generalisation, and interpretability in nonlinear structural dynamics.

Technical Skills

Programming Languages: Python, C++, MATLAB, JavaScript, Julia, C

Spoken Languages: Arabic, English, French, Hebrew

Software & Tools: NumPy, pandas, PyTorch, JAX, scikit-learn, SolidWorks, Inventor, Shapr3D, Abaqus

Developer Tools: VS Code, GitHub, Docker, GCP, pybind11

Certificates: Advanced Learning Algorithms (Coursera)