Home

E-mail: saum.bits@gmail.com

Google Scholar: link

Linkedin: https://www.linkedin.com/in/saumitra-mishra-phd-98082529/

About me

I am an Executive Director at JP Morgan Chase & Co. where I am a member of the AI research team and the research director of the Explainable AI Centre of Excellence.

Earlier, I was a research associate (postdoctoral researcher) at the Alan Turing Institute where I was a member of the Finance and Economics research programme. I worked on interpretable machine learning in the context of financial and computer vision applications. Interpretable machine learning involves approaches to understand the behaviour of machine learning models. I am highly motivated by the idea that in addition to high predictive power a machine learning model should also possess other highly important traits, like interpretability (e.g., provides a way to explain it's predictions), robustness (e.g., to adversarial perturbations), fairness (e.g., generates unbiased predictions). I was also a member of the DataSig project that aims to develop and employ mathematical tools (based on Rough Path Theory) for processing complex data streams.  

Earlier, I completed my PhD in machine learning at Queen Mary University of London, where I was a member of the Centre for Digital Music research group and the Machine Listening lab. My research involved developing post-hoc interpretable machine learning methods to understand the global and local behaviour of machine listening models that analyse audio. Some key use cases I worked on in my research include singing voice detection, automatic spoofing detection, and environmental sound classification.

Before my PhD, I worked at Samsung R&D Institute India, Bangalore for around seven years, where I contributed to several projects that fall under the domain of machine learning, audio signal processing and embedded systems. For more details please refer to my CV.

Research interests

News