Conducted a 6-month independent research project within a PhD-led program, supervised by a research director, on treatment outcome prediction from multimodal medical data (CT scans, clinical and genomic features). Built an end-to-end ML/DL pipeline with classical baselines and custom 2D/3D CNNs, and evaluated early and late multimodal fusion, including a novel CNN-based approach using radiomic parametric maps. Delivered a rigorous and reproducible benchmark on the largest public dataset in this setting, addressing key limitations of prior work.