Manuel Ciosici
Machine Learning Engineer at Apple
Los Angeles, California, United States
Summary
🤩
Rockstar🎓
Top SchoolManuel Ciosici is a Machine Learning Engineer and NLP researcher with 15 years of experience applying advanced ML to text understanding, now working at Apple from Los Angeles. He combines a strong academic foundation—a PhD from Aarhus University and multiple postdoctoral roles—with hands-on engineering experience contributing to major open-source projects like Hugging Face Transformers and DeepSpeed. His contributions include integrating 8-bit optimizer support (bitsandbytes) and practical code improvements that boost memory efficiency and logging reliability, showing a knack for production-focused research. Manuel’s background spans research, software development, and optimization at institutions such as the Information Sciences Institute, demonstrating an ability to bridge research prototypes into scalable tooling. Outside work he recharges outdoors, reflecting a practical, curious mindset that informs both experimentation and robust system design.
14 years of coding experience
9 years of employment as a software developer
Doctor of Philosophy - PhD, Doctor of Philosophy - PhD at Aarhus UniversityAP Degree, AP Degree at Erhvervsakademi Aarhus | Business Academy AarhusMaster’s Degree, Master’s Degree at Aarhus UniversitetBachelor's degree, Bachelor's degree at Aarhus Universitet / Aarhus University
Github Skills (18)
transformers10
pytorch10
python10
machine-learning10
model-optimization10
adam10
nlp10
testing9
deep-learning9
gpu8
f-string7
compression5
data-parallelism5
data-parallel5
parallelization5
Programming languages (6)
Github contributions (5)
deepspeedai/DeepSpeed
Sep 2021 - Jul 2022
DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.
Role in this project:
ML Engineer
ML EngineerContributions:5 reviews, 10 commits, 8 PRs in 9 months
Contributions summary:Manuel contributed to the DeepSpeed repository by addressing several minor issues, including fixing typos in documentation and code comments. They also replaced calls to `print()` with `logger.info()` for improved logging. Furthermore, the user updated code to use f-strings and modified a file related to parameter partitioning, demonstrating an understanding of the library's internal workings.
billion-parametersfinetuningtrainingmixture-of-expertszero
huggingface/transformers
Jan 2022 - Jul 2022
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:19 reviews, 5 commits, 7 PRs in 6 months
Contributions summary:Manuel contributed to the optimization and extension of the Hugging Face Transformers library, specifically by implementing and testing support for the bitsandbytes (bnb) library for 8-bit AdamW optimization. This involved integrating the bnb library into the training process, adding related tests, and verifying memory utilization improvements. Furthermore, the user added and modified optimizer options within the training arguments. The user also addressed documentation and code style.
pythonbertspeech-recognitionstate-of-the-artflax
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.