Ali Safaya
AI Research Scientist at HyperbeeAI
Istanbul, Istanbul, Turkey
Summary
👤
Senior🎓
Top SchoolAli Safaya is an AI research scientist and PhD candidate in Computer Science at Koç University with nine years of experience building NLP and speech systems for industry and research. He has worked repeatedly at HyperbeeAI and Codeway developing production-ready large language model and conversational AI solutions, and earlier contributed to ERC-funded research as a project engineer. An active open-source contributor, he added WavLM support and Arabic text-normalization fixes to the popular SpeechBrain PyTorch toolkit, demonstrating attention to real-world multilingual and numerical stability issues. Combining deep academic training with hands-on ML engineering, he excels at taking speech and language models from research prototypes to robust deployments.
9 years of coding experience
9 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Koç UniversityBachelor's degree, Computer Engineering, 3.89/4.00, Bachelor's degree, Computer Engineering, 3.89/4.00 at Sakarya Üniversitesi
Turkish, English, Arabic
Stackoverflow
56reputation
34kreached
5answers
1question
Github Skills (17)
transformers10
pytorch10
audio-processing10
speech-recognition10
nlp10
python9
speech-to-text9
huggingface-transformers9
deep-learning8
asr8
machine-learning8
eclipse6
logging6
operator6
dynamic-allocation6
Programming languages (6)
Github contributions (5)
speechbrain/speechbrain
Feb 2022 - Mar 2022
A PyTorch-based Speech Toolkit
Role in this project:
ML Engineer
ML EngineerContributions:7 commits, 4 PRs, 1 comment in 23 days
Contributions summary:Ali primarily contributed to the integration and support of pre-trained speech models within the SpeechBrain framework. This involved adding support for the WavLM model, alongside existing Wav2Vec2 and Hubert models. Furthermore, the user made modifications to language processing components by correcting Arabic character normalization rules and adding a missing character to improve text processing for the Arabic language. They also addressed potential numerical instability issues within the VAD energy calculation.
voice-recognitionasrspeech-recognitionspeech-separationspeaker-verification
alisafaya/SHA-RNN.jl
Feb 2020 - Nov 2020
Contributions:32 commits, 28 pushes, 1 branch in 9 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.