Partner Spotlight: Building AI for People, with David Brunner
Published Aug 31 2023 10:13 AM 1,546 Views


In this new edition of our Partner Spotlight series, we continue highlighting partners at the forefront of app innovation on the commercial marketplace. Throughout the series, we will be telling the unique stories of partners who are leading the way with AI in app development, who are building using multiple Microsoft products, and who are publishing transactable applications on the marketplace. In this article, I sat down with ModuleQ's David Brunner to learn more about their story and partner journey.


About David Brunner: Brunner founded ModuleQ in 2011 to pursue his vision for People-Facing AI that empowers professionals by understanding their business priorities and helping them work smarter. He received his PhD in Information, Technology & Management at Harvard University, where he studied digital transformation of business. Brunner holds a BS in Computer Science from Stanford, where he studied the design of data fusion systems with AI pioneer Professor Edward Feigenbaum. Brunner has peer-reviewed publications in artificial intelligence and operations management and received U.S. patents for ModuleQ’s Personal Data Fusion AI technology.


[JR]: Can you provide an overview of your application and its primary functionalities on the commercial marketplace?

[DB]: ModuleQ is a provider of applied People-Facing AI solutions that improve employee engagement and customer experience. People-Facing AI integrates with Microsoft Teams, the M365 Graph, and other business systems to learn each employee’s current customer priorities and predictively deliver relevant insights such as pre-meeting intelligence, deal alerts, research, and breaking news. These insights are optimized to improve performance of specific workflows in industry verticals such as investment banking, strategy consulting, wealth management, and sales trading. For example, Financial Sponsors Group bankers receive actionable intelligence about the M&A activity of their private equity clients. Regulated enterprises use People-Facing AI as an extensible, secure, compliant application to increase engagement with information sourced internally and from third-party providers, increasing the return on these information investments.


[JR]: Microsoft offers a variety of AI tools and services. Could you elaborate on which Microsoft AI tools you integrated into your app and how they have enhanced its capabilities?

[DB]: ModuleQ uses Azure OpenAI Service for fine-grained analysis of unstructured content, and Azure Machine Learning (ML) to implement ML models in production. Multiple Azure ML models power key capabilities of People-Facing AI. For example, when employees receive insights from ModuleQ AI in Microsoft Teams, they can provide feedback on which insights were Useful or Not Useful, allowing each user to develop their own personalized experience. By using Azure ML models to learn from this feedback loop, ModuleQ AI can more accurately predict the insights—research, marketing collateral, news, etc.—that will be most useful to each employee.


[JR]: In terms of user experience, how has the integration of AI tools positively impacted your application? Are there any notable features or functionalities that were made possible through AI?

[DB]: Yes, absolutely. One of our core use cases is to deliver personalized news updates that recommend articles curated to the interests and priorities of each employee.  Recommending news articles is a notoriously challenging problem, because there are so many different aspects to establishing the relevance of an article to the business context of a specific employee. ModuleQ AI uses multiple third-party natural language processing (NLP) models to perform basic metadata tagging of news articles, including technology from the London Stock Exchange Group (LSEG). In addition, we have trained our own proprietary classifiers to tune the news recommendations. We use Azure ML to deploy and manage these proprietary models. For example, one model scores news based on “newsworthiness.” This model filters out low-value articles, reducing information overload to help employees focus on the most important news. Another model improves the accuracy of news recommendations by identifying blocks of text that are “boilerplate” and not meaningful to substance of the article.


[JR]: How has ISV Success supported you in your journey to publishing on the marketplace?

[DB]: ISV Success plays an important role in ModuleQ’s efforts to develop the expertise of our engineers and mature our secure, scalable, enterprise-grade infrastructure on Azure. As an AI solution on Azure, it’s crucial for ModuleQ’s team to have comprehensive, cutting-edge Azure expertise. To date, ModuleQ engineers have achieved over 100 Azure certifications.  By co-investing with ModuleQ in these certification programs, ISV Success has helped ModuleQ accelerate and scale our Azure expertise. In addition, by co-investing with ModuleQ in Azure cloud infrastructure resources, ISV Success has supported our growth and maturity on the platform.


[JR]: Looking ahead, how do you envision further leveraging Microsoft's AI tools to enhance your application's features or expand its offerings in the marketplace?

[DB]: One of the most powerful Microsoft AI tools is the Azure OpenAI Service, which makes available the cutting-edge GPT models from OpenAI. As our customers enable Azure OpenAI Service within their own tenants, this will enable ModuleQ to perform fine-grained analysis of our customers’ documents and messages, improving the relevance and value of our insights while ensuring privacy and security. In our research lab, we have found that GPT queries can be used to identify content that is likely to be of interest to professionals seeking information about specific situations such as legal proceedings or financial transactions that have particular distinguishing characteristics. This content is often highly valuable, but difficult to identify using traditional natural language processing technologies.


[JR]: MLOps is an important priority for AI companies.  How have Microsoft’s AI tools helped your organization address MLOps challenges?

[DB]: Azure ML provides a robust toolset for distribution, versioning, upgrading, and monitoring of ML models across multiple customer installations. This has been extremely useful for ModuleQ, because of the decentralized ML architecture that we have implemented to meet our customers’ security and compliance needs. Each enterprise customer has their own instance of People-Facing AI running in their private Azure tenant.  Learning from user feedback takes place entirely inside the customer Azure tenant, eliminating the need for sensitive data to leave the customer’s governance perimeter and facilitating risk-acceptance of the solution. This architecture requires deploying ML models into each customer installation, making MLOps challenges even more complex.  Azure ML has been a great solution to these challenges.


[JR]: Privacy and security are crucial for AI applications in the enterprise, and are key tenets of Microsoft’s Principles for Responsible AI. How have Microsoft’s AI tools enabled your application to meet the security and privacy needs of your customers?

[DB]: By providing a mechanism for secure implementation of ML models in our customers’ Azure tenants, AzureML has become an important component of our secure AI stack. First and foremost, by using Azure ML in the customer tenant for the deployment of our ML models, we can eliminate the need for sensitive customer information to leave their governance perimeter. For additional security, Azure ML supports private endpoint connections within the customer VNet, so our models are not exposed to the Internet.


Join the marketplace community today! Just click "join" on the upper right corner of our marketplace community page. You can also subscribe to the community to stay updated on the latest stories of how these inspiring leaders carved their career paths, what lessons they learned along the way, and more. 









Version history
Last update:
‎Sep 19 2023 01:36 PM
Updated by: