Samyadeep Basu is a Research Scientist at Adobe Research, where he works on multimodal language models, vision-language models, and model controllability, serving as a key researcher on the post-training stack for small language models and vision-language models — shipping models that power grounding and retrieval use-cases across Adobe products. He completed his CS PhD at the University of Maryland, College Park with Soheil Feizi in the Center for Machine Learning, and prior to that spent two years as an Applied Scientist at Microsoft AI, with research stints at Microsoft Research (Cambridge and Redmond) and earlier internships at Adobe. His research sits at the intersection of understanding and control: how knowledge is stored and transferred inside multimodal models, and how to steer or edit those models with minimal intervention. Among his recent contributions is SliderEdit, a framework for continuous image editing with fine-grained, interpretable instruction control that disentangles individual sub-instructions in a multi-part edit prompt and exposes each as a globally trained slider, enabling smooth adjustment of edit strengths and compositional control across diverse editing scenarios — work that was accepted as an Oral at CVPR 2026. His broader research portfolio spans mechanistic interpretability and model editing for multimodal language models, with publications at NeurIPS, ICLR, COLM, EMNLP, and CVPR.