Encord

Encord is a platform for data-centric computer vision, offering tools to help machine learning engineers and data scientists improve training data quality and boost model performance. Founded in 2020, Encord provides solutions for annotating, managing, and evaluating training data for computer vision applications. The company's flagship product is Encord Active, an open-source toolkit that enables users to understand and enhance their training data quality, helping to bridge the gap between proof-of-concept models and those capable of performing reliably in real-world scenarios.

Encord Active utilizes a quality metrics approach, which involves the automatic calculation of characteristics of images, labels, model predictions, and metadata. This allows machine learning teams to find unknown failure modes in datasets, inspect data balance, identify outliers, and create actionable end-to-end active learning workflows. The platform aims to address the "production gap" that often hinders the widespread adoption of AI in critical applications such as self-driving cars and diagnostic medical models.

In addition to Encord Active, the company offers Encord Annotate, a collaborative data labeling platform designed to streamline the annotation process for various data types, including medical imaging. Encord's solutions have been employed by leading healthcare institutions to improve efficiency in tasks such as annotating pre-cancerous polyp videos and processing medical images.

Key customers and partnerships

Encord has established partnerships with several prominent healthcare institutions. King's College London utilized Encord's platform to annotate pre-cancerous polyp videos, resulting in a 6.4x increase in efficiency and automating 97% of labels. This improvement made the most expensive clinicians 16 times more efficient at labeling medical images. Additionally, Encord has worked with Memorial Sloan Kettering Cancer Center and Stanford Medical Centre, where their solutions reduced experiment duration by 80% and enabled the processing of three times more images.