UnBIAS is a cutting-edge text analysis and debiasing toolkit that assesses and corrects biases in textual data. It was built to tackle the inherent biases in AI systems and promote the ethical use of LLMs. UnBIAS addresses the urgent need for accurate and unbiased information dissemination in today’s digital age and promotes fairness and ethical AI practices.

A new interpretability framework for generative models, providing a user-friendly software interface for interacting with large-scale models spread out over multi-GPU and multi-node setups. FlexModel contributes to making advanced AI research more inclusive and universally approachable, lowering barriers and increasing their safety and trustworthiness.

CyclOps is a set of evaluation and monitoring tools that health organizations can use to develop and evaluate sophisticated machine learning models in clinical settings across time, locations, and cohorts.

The Principles in Action Playbook, developed by Vector’s AI Engineering team, contains guidance, examples, and tips to help those building products that leverage AI to do so responsibly. It was created to help professionals on the front lines – entrepreneurs, product managers, designers, tech leads, and domain experts – navigate the complexities of AI product development.

A foundational model from Vector’s AI Engineering team to optimize model deployment on-demand, empowering researchers and industry partners who may not have technical knowledge by allowing them to dive right in and begin experimenting with foundation models. A key enabler to research and adoption, the 2022-23 year saw almost 100 users from Vector’s community use Kaleidoscope over 700,000 times.

FLorist is a platform to launch and monitor Federated Learning (FL) training jobs. Its goal is to bridge the gap between state-of-the-art FL algorithm implementations and their applications by providing a system to easily kick off, orchestrate, manage, collect, and summarize the results of FL4Health training jobs.

FL4Health is a flexible, modular, and easy to use library to facilitate federated learning (FL) research and development in healthcare settings. It aims to make state-of-the-art FL methods accessible and practical, These include personalized algorithms meant to tackle difficult settings such as heterogeneous data distributions which are common in clinical settings.