Trubrics Beta Release

**该项目的目标是开发我们的人工智能保证解决方案和业务Trubrics，用于测试版。**Trubrics增加了业务和数据科学团队之间的协作和信任。这增加了参与人工智能项目的人员的多样性——增加了人工智能系统的可靠性，并减少了偏见。企业内部的机器学习（ML）项目是为了支持业务团队/领域专家做出关键业务决策而开发的。然而，由于缺乏业务团队的信任，企业正在努力增加ML项目的采用。影响ML计划的整体ROI。这是由于缺乏技术理解造成的，机器学习本身就存在风险：*机器学习模型难以解释*性能可能会随着时间的推移而降低*没有可靠的跟踪记录人工智能法规的引入，如欧盟人工智能法案，增加了人工智能项目的合规风险。这些法规的重点是降低“高风险”用例的风险，如信用评分，影响英国的关键行业，如金融服务和制药。高风险模型必须与关键需求保持一致，包括高质量的数据、适当的文档、透明度、人为监督，以及对准确性、稳健性和偏差的测试。Trubrics将提供一个解决方案，以保持领先于这些法规。Trubrics为业务团队提交ML模型的定性和定量反馈提供了一个解决方案。数据科学家使用这些反馈来改进和验证模型。Trubrics还提供了开发ML验证（测试）“金标准”的解决方案。这将使验证过程自动化，并确保验证是可重复的，并记录下来供审计使用。验证包括数据质量；建模性能、健壮性和偏差测试，以及自定义业务验证。在过去的10个月里，联合创始人自费开发了Trubrics MVP。包括一个SDK来收集用户反馈，并创建验证；以及一个软件即服务（SaaS）平台，用于存储、管理和可视化反馈和验证。一个早期访问计划目前正在进行中，以测试解决方案的初始用户。用户看到了解决方案的价值，但是**需要技术和商业开发的关键部分才能将解决方案和业务带到Beta版本。**该项目旨在完成测试版所需的开发，并**将产品推向市场

**The goal of this project is to develop our AI Assurance solution and business, Trubrics, for a Beta release.**Trubrics increases collaboration and trust between business and data science teams. This increases the diversity of the people involved in AI projects - increasing the reliability of the AI system, and reduces bias.Machine learning (ML) projects within businesses are developed to support business teams/domain experts with key business decisions. However, businesses are struggling to increase the adoption of ML projects, due to a lack of trust from the business teams. Impacting the overall ROI of ML initiatives.This is caused by a lack of technical understanding, and ML being inherently risky:* ML models are difficult to explain* Performance can degrade overtime* No proven track recordThe introduction of AI regulations, such as the EU AI Act, increases the compliance risk of AI projects. These regulations are focused on reducing the risk of 'high-risk' use cases, such as credit scoring - impacting key UK industries, such as Financial Services and Pharmaceutical.High-risk models must align with key requirements, including high-quality data, appropriate documentation, transparency, human oversight, and testing for accuracy, robustness and bias. Trubrics will provide a solution to stay ahead of these regulations.Trubrics provides a solution for business teams to submit qualitative and quantitative feedback for ML models. This feedback is used by data scientists to improve and validate the models.Trubrics also provides a solution to develop a 'gold standard' of ML validations (tests). This automates the validation process, and ensures validations are repeatable and logged for audit. Validations include data quality; model performance, robustness and bias tests, and custom business validations.Over the last 10 months, the Co-Founders have self-funded the development of the Trubrics MVP. Including an SDK to collect user feedback, and create validations; and a software-as-a-service (SaaS) platform to store, manage and visualise both feedback and validations.An Early Access program is currently underway, to test the solution with initial users. The users see value in the solution, however **there are key pieces of technical and commercial development required to take the solution and business to Beta release.**This project aims to complete the development required for the Beta release, and **take the product to market.**