Roshan Sharma
Research Scientist at Google DeepMind
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolRoshan Sharma is a Senior Research Scientist at Google DeepMind specializing in audio post-training and agentic audio interactions, where he also leads technical efforts on Audio Tool-Use for the Gemini audio team. He holds a PhD in Speech and Language Technologies from Carnegie Mellon and brings seven years of experience across speech recognition, spoken language understanding, and multimodal modeling. Roshan has practical impact in open-source speech tooling, contributing recipe fixes and backend improvements to the widely used ESPnet end-to-end speech toolkit. His background spans industry and academia—from Meta and Qualcomm internships to doctoral research—giving him a rare blend of production-minded engineering and deep research rigor. Based in New York, he’s particularly focused on making machines perform complex, human-like speech understanding tasks and improving the toolchains that enable those advances. An underappreciated strength is his hands-on work fixing real-world ASR recipes, bridging research prototypes and deployable pipelines.
7 years of coding experience
1 year of employment as a software developer
Bachelor of Technology (B.Tech.), Electrical, Electronics and Communications Engineering, A, Bachelor of Technology (B.Tech.), Electrical, Electronics and Communications Engineering, A at Amrita School of Engineering BangaloreDoctor of Philosophy - PhD, Electrical and Computer Engineering, Doctor of Philosophy - PhD, Electrical and Computer Engineering at Carnegie Mellon UniversityCertificate, Micro and Nano-Science, Certificate, Micro and Nano-Science at Institut polytechnique de Grenoble
English, French, Kannada, Tamil, Hindi
Github Skills (11)
sh10
data-handling10
script10
speech-recognition10
shell10
automatic-speech-recognition10
asr10
scripting10
computer-engineering8
python7
machine-learning6
Programming languages (3)
Github contributions (5)
espnet/espnet
Jun 2021 - Dec 2022
End-to-End Speech Processing Toolkit
Role in this project:
Back-end Developer
Back-end DeveloperContributions:14 reviews, 139 commits, 17 PRs in 1 year 5 months
Contributions summary:Roshan primarily contributed to the "espnet/espnet" repository by fixing and updating the code related to the SWBD (Switchboard) recipe. The changes involved modifications to scripts like `show_asr_result.sh`, and `data.sh` file within the `egs2/TEMPLATE/asr1` directory, indicating a focus on speech recognition result display and data handling. Further contributions include merges and recipe fixes, mainly in the egs2/swbd/asr1 directory, showing hands-on work with speech processing tools within the repository. The nature of the changes suggest the user was involved with improving the recipe.
speech-recognitionspeech-separationchainerspoken-language-understandingspeech-processing
DavidLBick/espnet-1
Mar 2022 - Feb 2025
End-to-End Speech Processing Toolkit
Contributions:2 PRs, 43 pushes in 2 years 11 months
end-to-endspeech-to-textspeech-recognitionspeech-synthesisautomatic-speech-recognition
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.