Summary
Roman Koshkin is a research scientist based in Tokyo with eight years of experience at the intersection of speech processing, machine translation, and biologically-inspired AI. He builds real-time, full-duplex speech-to-speech systems and multimodal human-AI interfaces, blending deep learning for NLP with insights from computational neuroscience. A PhD researcher and postdoc from OIST, he has practical experience deploying simultaneous translation models and prototyping EEG- and robot-control systems at industry labs and startups. Roman’s background in linguistics and cognitive science informs his work on latency-sensitive, brain-aligned models—he’s applied spiking networks, STDP and contrastive pretraining to bridge biological plausibility with production-grade speech translation.
8 years of coding experience
1 year of employment as a software developer
Doctor of Philosophy - PhD, Machine Learning and Computational Neuroscience, Doctor of Philosophy - PhD, Machine Learning and Computational Neuroscience at Okinawa Institute of Science and Technology
Master's degree (with distinction), Cognitive Sciences and Technologies, PSYCHOLOGY, 8.9 out of 10, Master's degree (with distinction), Cognitive Sciences and Technologies, PSYCHOLOGY, 8.9 out of 10 at State University — Higher School of Economics
Bachelor's degree (with distinction), Linguistics, 4.9 out of 5, Bachelor's degree (with distinction), Linguistics, 4.9 out of 5 at VUMO University
Linguistics, Linguistics at York St. John University
English, Russian, French