phd-research

PhD research: predicting flood risk and flood depth with self-supervised learning on SAR imagery. Remote sensing, physics-informed representation learning and computational hydrology, aimed at insurance risk in data-scarce regions. University of the Witwatersrand.

A Claude Code-native research operating system: 28 specialized research subagents, multi-agent slash-command panels, cited domain knowledge notes, and living memory.

Research hub — interactive computational research on complex systems, emergence, and collective behavior. PhD research identity of Anthony Seumal.

AI-powered medical image analysis using Google's MedGemma 1.5 (4B) — generate structured radiology reports from X-rays, MRI, CT, pathology slides, and more.

SIMB Research Started Summer 2026

Detecção Precoce de Estudantes em Risco de Evasão Usando Dados Administrativos e Aprendizagem de Máquina

Interactive scrollytelling on GenAI workplace adoption, simulated through Schelling's Segregation model and Granovetter's Threshold Theory — when does individual AI use become a collective norm? Emergence Lab research.

Bioinformatics analyses using R, including MSc and PhD projects

Detector-agnostic ML classifier for hit/non-hit identification in Serial Femtosecond X-ray Crystallography (SFX), generalizing across AGIPD, JUNGFRAU 4M, ePix10k, and Eiger4M detectors.

Bioinformatics analyses using R, including MSc and PhD projects.

Using clustering to define roost sites

PhD Research Framework for Data-Driven Modeling of Elastic-Plastic Material Behavior. A comprehensive library providing tools for defining, training, and validating neural network-based constitutive models against simulation data.

NBDA analyses of social information in vultures

Synthesis of High-Entropy Alloys via cold plasma: waveform analysis, FFT processing, Townsend discharge modelling, and mixing entropy calculations.

Initial exploration of vulture data

Native Android (Kotlin + Jetpack Compose) client for XAI-FLOWS — real-time flood-risk monitoring with a background foreground service, SHAP explainability, and full account auth. PhD research project.

RL-powered test prioritization (PPO + MaskablePPO) with an honest engineering writeup: found the agent scoring worse than random, root-caused and fixed it, validated via a 5-seed sweep. Includes a genuine multi-agent AutoGen negotiation layer.

Simulation study of cache architecture tradeoffs under concurrency — partitioned vs LRU vs client affinity, with trace-driven evaluation on Twitter cache workloads

GNN-based analysis of brain functional connectivity from EEG — cognitive workload classification and ADHD detection using GCN/GAT on 19-channel brain networks, with an interactive Dash dashboard.

Comparative simulation study of TMLE and C-TMLE estimators for causal inference in high-dimensional observational data · R