Bassel El Mabsout

Co-PI, representing Neobotics (PI: Renato Mancuso)
Proposal to the NSF Translation to Practice program (NSF 25-540). The project, NeoRacer, turns NSF-catalyzed lab research into mass-produced, turnkey kits for autonomous-vehicle research and education, manufactured by Neobotics. Submitted May 2026, under review.

PIs: Benjamin Lubin, Marco Caccamo, Sandro Pinto, Renato Mancuso
Selected proposal and Best Presentation Award winner at the GenZero Workshop. Contributed to the proposal and led the prototype demonstration across Boston University, TU Munich, and University of Minho, for a framework separating data ownership, processing, and usage concerns.

PI: Renato Mancuso
Accepted Letter of Interest for a plug-and-play confidential computing solution: secure application execution on traditional server hardware, maintaining security even when the server is compromised.

PIs: Renato Mancuso, Sabrina Neuman
NSF-medium proposal on neural network control for under-instrumented limbed robots, HW/SW architectures for energy-efficient control, and learned runtime adaptation on constrained platforms.

PIs: Renato Mancuso, Marco Caccamo
NSF-medium proposal on eco-friendly UAV design for agriculture: solar power with neural networks maximizing power-efficient control under formal safety bounds.

PIs: Renato Mancuso, Chuchu Fan
Joint proposal with MIT on certifiably safe neural network controllers that adapt to substantial system damage while maintaining provable safety guarantees.

A novel strategy for enhancing the robustness of reinforcement learning agents in crossing the sim-to-real gap, achieving near-50% reduction in power consumption while maintaining controllable, stable flight.

A population-based hyperparameter optimization method that adaptively controls hyperparameter selection via a normalized fitness function, outperforming schedules by up to 18%.

A method for efficiently computing the gradient of Boolean functions by constructing computational graphs effectively.

Learned controllers using bounded Lyapunov functions for maintaining stability under multiple complex dynamical systems.

A risk-aware path planning framework using 2D-Gaussian mixtures for real-time sampling and optimization on F1tenth vehicles.

Survey exploring safe robot learning, investigating methods that balance high-performance control with rigorous safety constraints.

Built Generative Adversarial Networks modeling the distribution of stochastic system trajectories for safer learned model-based control.

Worked on privacy-preserving preferential ride-sharing using secure multi-party computation with BU's SAIL and Honda.

Kaggle competition project achieving highest score (AUC 0.92) for predicting seizure activity in epileptic patients.

Programming languages blog exploring geometry, automatic differentiation, and dependently typed vector construction. Featured on Haskell Weekly.