Summary
Soham Deshmukh is a research scientist specializing in speech and language processing with nine years of experience building audio-centric ML systems. He has driven production-grade solutions at Microsoft—spanning speech enhancement, source separation, ASR adaptation, and multimodal pretraining that power features across Edge, Teams, and Azure—and now focuses on audio and language at Sesame. His academic work at Carnegie Mellon produced unsupervised vocal fold motion modeling and weakly supervised sound event detection, with papers in ICASSP and INTERSPEECH, reflecting a strong research-to-product trajectory. Comfortable across research and applied roles, he blends rigorous signal-processing insight with pragmatic engineering to ship robust models. Colleagues note his knack for turning complex audio physics into deployable ML components that improve real-world communication experiences.
8 years of coding experience
8 years of employment as a software developer
Bachelor of Technology Electronics Engineering, Bachelor of Technology Electronics Engineering at Veermata Jijabai Technological Institute (VJTI)
Doctor of Philosophy - PhD Electrical and Computer Engineering, Doctor of Philosophy - PhD Electrical and Computer Engineering at Carnegie Mellon University