Projects
Research and engineering projects, each with its problem, methodology and outcomes. Quantitative results are reported only where they appear in a published paper or preprint.
Medical imaging · Vision-language
Generating expert-level radiology reports directly from volumetric chest CT by adapting a RadImageNet-pretrained ResNet-50 to 3D data with axial-coronal-sagittal convolutions and decoding with GPT-2 through cross-attention, trained on a single 24 GB GPU and a fraction of the CT-RATE dataset.
Lead author · MSc thesis research, AIMS South Africa / UWC (2024)
AI for Science · Scientific machine learning
Retrieving vegetation biophysical variables from Sentinel-2 imagery by inverting the PROSAIL radiative transfer model with neural networks that use the physical model as a differentiable decoder, trained only on simulated data.
AI Research Engineer Fellow, InstaDeep (Nov 2024 – May 2025) · Co-author of the preprint
Generative AI · Reasoning in complex systems
A vision for embedding large language models in the 6G radio access network control loop, with a formal framework aligned to the O-RAN architecture and results on policy expressiveness and convergence.
Co-author · Workshop paper, NeurIPS 2025 AI4NextG
Multimodal interaction · Embodied AI
Engineering real-time speech-to-speech pipelines, LLM reasoning and robot control for an embodied humanoid AI system at IQAI, with a focus on conversational latency, reliability and safe autonomous behaviour.
AI Research Engineer, IQAI (June 2025 – present)