SBIR Phase I: An Open Source Platform for Intelligent Virtual Assistants

这个小型企业创新研究(SBIR)第一阶段项目的更广泛影响/商业潜力将来自世界上第一个开源智能虚拟助手(IVA)平台的开发和发布。该平台将作为开放源码软件免费向世界提供，并将围绕着一系列商业产品，其中包括软件的部署和定制，以及软件即服务(SaaS)产品，使平台用户能够轻松地创建和部署IVA。除了IVA技术的商业应用，该平台的开源发布将有助于围绕IVA技术培育一个健康、活跃的开源社区，使其能够被用作机器学习研究平台，并加速IVA在非营利性教育等重要非商业环境中的采用，并改善残疾人的技术获取。这个小型企业创新研究(SBIR)第一阶段项目将解决在支持智能虚拟助理(IVA)技术所需的算法组件的复杂管道中无缝协调大量计算所涉及的重大计算机系统挑战。为IVA提供动力所需的计算量足够大，这样的软件系统可能会在由不同服务器类型和加速器平台的复杂生态系统组成的大型数据中心基础设施中运行。该项目将通过在IVA平台内建立一个用于监测作业执行情况的集中软件机制来解决这些挑战，以促进有效地将IVA计算中的子任务映射到数据中心的现有硬件资源。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project will arise from the development and release of the world's first open source Intelligent Virtual Assistant (IVA) platform. The platform will be made available to the world as open source software free of charge, and will be surrounded by a set of commercial offerings that include the deployment and customization of the software as well as software-as-a-service (SaaS) offerings that will allow users of the platform to painlessly create and deploy IVAs. Beyond the commercial applications of IVA technology, the open source release of the platform will help foster a healthy, active open source community around the IVA technology, allowing it to be leveraged as a machine learning research platform and accelerating IVA adoption in important non-commercial settings such as non-profit education and improving technological access for the disabled.This Small Business Innovation Research (SBIR) Phase I project will address the significant computer systems challenges involved in seamlessly orchestrating the massive amounts of computation among complex pipelines of algorithmic components required to power intelligent virtual assistant (IVA) technologies. The computation required to power IVAs is large enough that such software systems will likely run in large datacenter infrastructures comprised of complex ecosystems of different server types and accelerator platforms. This project will address these challenges by building a centralized software mechanism within the IVA platform for monitoring job execution to facilitate efficiently mapping sub-tasks within the IVA computation to the available hardware resources in the datacenter.