onsite
Data Engineer - Research
Data Engineer
The Data Engineer - Research will focus on scaling efficient distributed workloads and improving data efficiency for machine learning models. Key responsibilities include data cleaning, preprocessing, building data pipelines, and AWS resource management for groundbreaking projects like training large open language models.
About the role
About the Role
We are looking for a talented Data Engineer with a focus on scaling efficient distributed workloads. You will work alongside a growing multidisciplinary team of talented research scientists and machine learning engineers to improve and scale the efficiency within our models. In this role, you will contribute to groundbreaking projects such as training the largest open language models and be responsible for ensuring data is collected, processed and utilized in the right way.
Responsibilities
- Clean, normalize, and preprocess data in a scalable, parallelizable way to prepare it for ingestion into our machine learning model training pipelines while ensuring of data quality
- Building and maintaining highly scalable distributed workloads
- Build data pipelines to ingest and process data (e.g. images and text) for feeding into ML models
- AWS Resource Management
- Keep up-to-date with methods regarding how to improve data quality and/or curate data for Image, Video, LLMs etc
Qualifications
- Proven background within large scale distributed workloads
- Experience with large scale data loading for machine learning training runs
- Experience with cloud storage and file systems. AWS (S3) is strongly preferred, but open to other cloud platforms
- Experience with Python + Pytorch
- Experience with multiprocessing and multithreading python workloads
- Excellent communication skills to effectively collaborate with users, solve issues, and provide guidance
- Attention to detail and the ability to document processes and solutions effectively
- Strong interest in Generative AI
- Experience working with Machine Learning projects and ideally some Deep learning / Comp Vision knowledge
- Experience with dataloading stack (webdataset, torchdata, fsspec, AIstore) and parallel dataframe manipulation using Pyspark/Ray is a plus point