March 2025 Summaries

The creation of large language models is like the advent of electricity, new and paradigm-shifting, but not well understood in terms of how it can be used. To identify use cases for this technology, it's helpful to start with an existing problem and consider how generative AI can solve it. The long and labor-intensive process of underwriting loans and mortgages is a prime example, where LLMs can speed up the process to half the time it would take manually. An implementation using a Compound AI approach combines many different AI components, data sources, and integrations to power mission-critical applications. A digital assistant called the "AI underwriter" works alongside human analysts throughout the underwriting workflow, automating labor-intensive aspects of underwriting and augmenting their capabilities. The copilot acts as connective tissue between scattered loan information and the analyst's decision-making process, handling tasks such as document processing, extracting critical data points, analyzing relationships, answering specific queries, and populating standardized sections of credit memos. By automating these labor-intensive aspects of underwriting, the copilot reduces the workload of analysts, allowing them to focus on higher-value risk assessment and decision-making. The implementation is technically implemented through a Compound AI approach, which requires a customized end-to-end solution with advanced AI models, data sources, and integrations. The power of the AI underwriter lies in its ability to combine specific details from multiple information sources to paint a nuanced picture of each deal. Deploying Compound AI has traditionally required a large team of experts, but AI orchestration frameworks and tools have changed that, making it easier for organizations to build quickly and within budget.

Deepset has announced its Custom AI Agent Solution Architecture, designed to enable enterprises to deploy AI agents both securely and efficiently across cloud and on-premises environments. This architecture combines deepset's LLM orchestration capabilities and the Haystack open-source framework with NVIDIA AI Enterprise software, enhancing organizations' ability to build and manage AI agents while maintaining operational control and data security. The solution supports enterprise-grade security and compliance and features NVIDIA's Triton Inference Server and NeMo microservices for optimized performance. Organizations like Airbus, OakNorth, and The Economist already utilize deepset's platform to create custom AI applications suited to their needs, ensuring advanced orchestration, infrastructure sovereignty, and performance optimization. The integration of NVIDIA AI Enterprise enhances operational capabilities with improved deployment speed and efficiency, enabling comprehensive and faster insights for businesses, as demonstrated by YPulse's experience. Attendees at GTC 2025 will have the opportunity to explore the deepset-NVIDIA AI Agent Solution and its potential to accelerate AI adoption.

deepset has announced significant advancements in its collaboration with Amazon Web Services (AWS) through the AWS Partner Network, including a Strategic Collaboration Agreement, the availability of the deepset AI Platform on the AWS Marketplace, integration with Amazon Bedrock, and recognition for the AWS Generative AI Competency. These developments aim to strengthen deepset's role as an AWS partner, enabling businesses to adopt AI rapidly by leveraging use cases like AI agents and semantic search. Customers such as credX and MANZ have reported notable improvements in efficiency and research capabilities by utilizing deepset's AI solutions integrated with AWS's infrastructure. The partnership facilitates the accelerated deployment of custom AI solutions, offering enterprises enhanced procurement and performance options through AWS Marketplace, while deepset's recognition as a Gartner Cool Vendor and AWS Generative AI Competency award underscore its leadership in AI engineering.

Deepset and PwC have joined forces to expedite the adoption of generative AI in enterprises by offering customizable agents and applications that cater to specific business needs. This collaboration leverages deepset's enterprise AI orchestration technology and PwC's strategic advisory services to cut implementation time to weeks while allowing for integration of new AI innovations. The deepset AI Platform, which includes the Haystack open-source LLM orchestration framework, helps enterprises rapidly build, test, and deploy modular AI solutions by selecting suitable data, large language models, integrations, and deployment options, thus enhancing AI accuracy and adoption. This system complements PwC's partner ecosystem, maximizing AI's business impact and ensuring enterprises remain adaptable and free from vendor lock-in. The partnership aims to future-proof AI strategies, with deepset customers reportedly achieving 40% efficiency gains and a fivefold return on investment in critical use cases. Mark Ghannam of deepset highlights the partnership's potential in tackling complex AI challenges and delivering significant business impact by combining cutting-edge technology with strategic leadership.