What AI features are available for researchers?

Jonathan Ozeran: Tempus Lens is our cloud-based data analytics platform, harnessing the power of multimodal data to assist our life sciences partners and researchers. The goal of Lens is to simplify the process of stratifying patient cohorts, making it as straightforward and swift as possible. The platform boasts a suite of basic and advanced filters, designed to streamline the search process across Tempus’ extensive dataset of over 8 million de-identified patient records.

One standout generative AI feature of Lens is its cohort-building functionality, which is currently in Beta. When exploring complex queries, for example, identifying patients with triple-negative breast cancer treated with specific drugs and carrying certain genetic mutations, Lens employs gen AI agents to dissect the query and apply the appropriate filters. This system helps to distinguish between different assays and mutations, ensuring that only clinically significant results are considered. What was once a time-consuming task now returns results in mere seconds. This rapid turnaround is transformative for researchers who can move from hypothesis to sizing a cohort of interest almost instantaneously.

Upon applying the filters, Lens doesn’t just stop at providing a number; it offers a visualization interface with detailed information, including demographics and genetic variants. For a deeper dive, it can display additional analyses like disease response derived from structured and abstracted clinical data from patient records. Lens has lowered the barrier to entry for using real-world data. Previously, navigating analytical tools might have required extensive training and familiarity with medical nomenclature. Now, thanks to the natural language capabilities of Lens’ cohort-building tool, it’s as easy as conversing in your own terminology.

While structured and molecular data are readily available, Lens also acknowledges the challenge of unstructured clinical notes. These documents often contain valuable insights, such as adverse events, symptoms, or specific details from imaging tests, which are not always captured in structured EMR data. Lens is evolving to incorporate these elements, making it an even more comprehensive tool for research and analysis.

How will Lens support the interrogation of unstructured data elements?

What approach does Tempus take to developing new AI capabilities for its products?

Jonathan Ozeran: We’ve developed an in-house version of Tempus One using an agentic workflow platform that we call “Agent Builder.” This proprietary, custom-built toolkit enables our team to construct additional agents that address the needs of our clients. One example is our Publications Agent, which sifts through our extensive collection of publications to quickly retrieve information such as authors, conference details, and publication dates. This process is streamlined through what we call ‘assemblies’ within the Agent Builder. The ‘assemblies’ manage the inputs, orchestration tasks, and outputs, culminating in a conversational interface that is both user-friendly and versatile.

Behind the scenes, Agent Builder allows us to have full programmatic control over how these agents function. We can specify the source of the documents, choose the model, and define the output. This level of customization is crucial as it informs how we architect applications and bring them to market.

Even more innovation occurs in the workflow within Agent Builder. Here, we assemble the building blocks of an agent, allowing us to craft queries and receive responses. For example, we can create an agent to query PubMed with a focus on ADCs in oncology, and the agent will think through the process, using the collective intelligence of all the agents we’ve built at Tempus. This technology isn’t just for show; it’s a practical tool that empowers our team members to ask questions and get answers quickly. Whether it’s summarizing research or creating an agent from scratch, Agent Builder provides a functional tool that can be used in seconds, demonstrating the agility and power of AI in advancing our work at Tempus.