Soumya Vanka

I'm a PhD researcher in Audio and Music Technology at Queen Mary University of London currently working on AI-assisted music production and mixing style transfer. I previously did research at the Centre for Digital Music (C4DM) working with Dr. George Fazekas and Jean-Baptiste Rolland.

My work focuses on developing intelligent systems for music mixing and production, particularly emphasising mixing style transfer. I've developed Diff-MST, a framework comprising a differentiable mixing console and transformer controller that can transfer mixing styles between songs while maintaining interpretability and control. This work addresses key challenges in automated music production by allowing for post-hoc adjustments and supporting variable numbers of input tracks.

Beyond technical development, I'm also interested in understanding how AI technology is being adopted in professional music production workflows. Through my research, I've investigated how different user groups - from amateurs to professionals - interact with AI mixing tools, and how communication and reference songs play crucial roles in the mixing process. This human-centered approach helps inform the development of more effective and user-friendly AI systems for music production.