Manmay Nakhashi
Machine Learning Researcher at Resemble AI
Pune, Maharashtra, India
Summary
🤩
Rockstar🎓
Top SchoolManmay Nakhashi is a Machine Learning Researcher based in Bengaluru with 10 years' experience building production speech and audio systems, currently advancing voice technologies at Resemble AI. He has deep hands-on expertise in TTS, speech synthesis, ASR and model optimization—delivering real-time, low-memory deployments on edge platforms and contributing to voice cloning, prosody modeling and cross-speaker emotion transfer. Manmay has applied these skills across startups and enterprise projects (Visteon, Dubdub.ai, ConcertAI) and improved inference performance in notable open-source TTS work such as integrating DeepSpeed and kv_cache optimizations into the popular Tortoise TTS codebase. Known for blending research and engineering, he repeatedly turns complex audio ML models into efficient, deployable systems for constrained hardware and multilingual applications.
10 years of coding experience
7 years of employment as a software developer
Post Graduate diploma, System software development, Post Graduate diploma, System software development at centre for development of advance computing(CDAC)Bachelor of Engineering (B.E.), Computer Science, Bachelor of Engineering (B.E.), Computer Science at Gujarat technological university
Stackoverflow
1reputation
0reached
1answer
0questions
Github Skills (10)
pytorch10
machine-learning10
text-to-speech10
deepspeed10
inference10
python10
model-optimization10
user-interface6
profiler6
tensorflow6
Programming languages (4)
Github contributions (5)
neonbjb/tortoise-tts
Jul 2023 - Dec 2024
A multi-voice TTS system trained with an emphasis on quality
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 35 PRs, 73 pushes in 1 year 5 months
Contributions summary:Manmay primarily focused on integrating and optimizing the DeepSpeed library for faster inference within the Tortoise TTS system. Their contributions included adding DeepSpeed support, addressing compatibility issues, and reverting changes due to identified crashes. Further work involved adding kv_cache functionality and refining model configurations for memory efficiency. The user also made minor adjustments to other aspects of the codebase, such as adding speaking rate control.
pytorchvoicequalityemphasistts
manmay-nakhashi/activity-recognation-based-on-posenet
Apr 2020 - Jul 2020
Contributions:4 commits, 7 pushes, 1 branch in 3 months
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.