Machine Learning Engineer, Data Quality

About the AI & Biometrics Team:

The AI & Biometrics team is building a biometric iris recognition system that can work reliably with more than a billion users and enables them to claim their free share of WLD. We use cutting-edge machine learning deployed on custom hardware to enable high-quality image acquisition, identification, and fraud prevention, all while requiring minimal user interaction. Our technology, coupled with privacy-preserving data collection, allows us to increase system performance and reduce model bias.

About the Opportunity:

This role is responsible for the data pipelines that fuel our machine learning engines. From dedicated field tests all over the world we receive millions of images monthly. Our ML models—especially the identifier models—require large high-quality datasets. To create datasets on such a scale, images need to be pre-processed and passed through our labelling services; this role is responsible for designing and building such pipelines to generate high-quality datasets.

In this role you will: 

• Design data pipelines for large scale data ingestion. This includes figuring out ways to store and process the data with robust features for filtering, pre-processing, and versioning.
• Build and refine custom data labeling services that directly influence the quality of our ML models.
• Work closely with other internal stakeholders to incorporate their data usage needs.
• Improve our data quality by deploying techniques like semi-supervised learning, human-in-the-loop machine learning, and fine-tuning with human feedback.

About You:

• Have industry experience as a Data Engineer, Machine Learning Engineer, or Data Scientist; you are comfortable with large amounts of data.
• Own problems end-to-end, and are willing to pick up whatever context is needed to get the job done.
• You enjoy working as part of a fast-moving team.
• A desire to dig into problems across the stack, whether networking issues, performance bottlenecks, memory leaks, or simply reading unfamiliar code to figure out where potential issues might exist.
• Have a strong belief in the crucial need of high-quality data for producing state-of-the-art machine learning systems.