Kritika Singh
Software Engineer at Facebook
Mountain View, California, United States
Summary
👤
Senior🎓
Top SchoolKritika Singh is a Software Engineer with 13 years of experience based in Mountain View, California, focused on speech recognition and ML systems. She has been at Facebook since 2015 and also conducts research at UC San Diego, bridging production engineering with academic inquiry. Her open-source contributions to Facebook Research's fairseq include improving CTC support, integrating wav2vec models, and hardening distributed training and checkpointing workflows. That blend of production-grade engineering and cutting-edge speech model development enables reliable deployment of advanced ASR components. Trained at IIT Kanpur and UC San Diego in computer science, she brings deep technical foundations and a knack for making research-ready models work at scale. An under-the-radar strength is her experience adapting pretraining pipelines to cooperate with external decoders, which smooths integration between research prototypes and real-world systems.
13 years of coding experience
University of California San DiegoIndian Institute of Technology KanpurDelhi Public School, Kalyanpur
Github Skills (10)
pytorch10
machine-learning10
artificial-intelligence10
python10
model-optimization9
optimization9
optimisation9
speech-recognition9
distributed-training9
nlp9
Programming languages (1)
Github contributions (5)
facebookresearch/fairseq
Apr 2019 - Feb 2021
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:
ML Engineer
ML EngineerContributions:15 commits in 1 year 10 months
Contributions summary:Kritika primarily contributed to the fairseq library, focusing on enhancements and bug fixes related to machine learning models. They implemented functionalities to improve the CTC (Connectionist Temporal Classification) criterion, enabling its compatibility with various encoder architectures. Additionally, the user worked on integrating wav2vec models, including adjustments for compatibility with external decoders and changes to the audio pretraining task. Furthermore, the user addressed issues related to distributed training and checkpoint averaging scripts.
pytorchnlpsequencepythontransformer-architecture
skritika/Sentiment-Analysis-using-Recursive-Autoencoders
Mar 2014 - Jan 2015
Contributions:8 commits, 3 pushes in 10 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.