Researcher in Deep Learning & Astroparticle Physics

§1Research Focus

My current work focuses on the core position reconstruction problem in extensive air shower (EAS) experiments. When a high-energy cosmic ray or gamma ray enters the atmosphere, it produces a cascade of secondary particles — an extensive air shower — whose footprint is sampled by ground-level detector arrays. Accurately determining where the shower axis intersects the detector plane is critical for energy estimation and gamma/hadron separation in observatories such as SWGO and HAWC.

I designed BiLayerEdgeConv, a dual-branch graph neural network that independently encodes electromagnetic and muonic signals from a dual-layer water Cherenkov detector (WCD) array. Each branch applies dynamic edge convolution over detector-level graphs, and the representations are fused through bidirectional cross-attention before regression. On Geant4/CORSIKA-simulated data, the model achieves a mean absolute error of 45 meters — a 20% improvement over single-layer baselines.

Broader interests include equivariant architectures for structured scientific data, physics-informed neural networks, and the application of geometric deep learning to detector reconstruction problems in astroparticle and high-energy physics.

§2Selected Publications

1. [1]Martinez Castro J.A., Misonia H., et al. “Suppressor Of Cytokine Signaling Proteins As Regulators Of Innate And Adaptive Immune Response.” Journal of Population Therapeutics and Clinical Pharmacology, Vol 30 No 18, 2023.
2. [2]Misonia H. “BiLayerEdgeConv: Dual-Branch Graph Neural Network for Extensive Air Shower Core Reconstruction on Water Cherenkov Detector Arrays.” MSc Thesis, CIC-IPN. In preparation.

§3Engineering

Seven years building production systems — financial management platforms for law firms, industrial audit implementations, educational platforms, wildlife crime awareness applications, fiscal management tools, hospital IT solutions, and hotel operations systems. Full-stack development across React, Vue.js, Node.js, Laravel, Spring Boot, and PostgreSQL, serving clients across the DR Congo.

§4Selected Clients

During my tenure at Kadea Software and as a freelance developer, I delivered production systems for clients across legal, industrial, healthcare, government, and education sectors — including USAID, Finasucre (Compagnie Sucrière de Kwilu-Ngongo), GoLegal, Clinique Marie Yvette, Ensemble pour la République, and Leon Hotel.

§5Contact

Open to: PhD positions, research collaborations, visiting appointments.