Updated March 5, 2026
0:00 Welcome to Colabira AI podcast brought to you by Colabira AI Research Labs and Carl Foundation. Today, we're taking a deep dive into Tempus AI, a company squarely focused on, AI enabled precision medicine. Yeah. And for this deep dive, our exploration is really grounded in, you know, recent information you've shared. We're looking at details about their electronic health record or EHR integration efforts, some new AI features they've introduced, and importantly, kind of digging into insights straight from their own descriptions of their technology and, well, the science behind it. 0:31 Absolutely. So our mission for this deep dive is pretty clear. We wanna unpack Tempest's fundamental approach. We wanna understand the core technology pillars they talk about. Explore how they apply this stuff across different medical fields. 0:46 And really see how they leverage artificial intelligence and frankly vast amounts of data to impact patient care, accelerate research, and improve drug development. And specifically, as you asked, we're gonna be looking closely at the technical methods they describe. And, you know, the results they report and the sources, what are they actually doing under the hood? What's the evidence it works? Okay. 1:07 Let's unpack this. Starting with that core concept they position themselves around, AI enabled precision medicine. What does that phrase truly imply in the context Tempest uses it? Good question. At its heart, it's really about using artificial intelligence. 1:21 You know, advanced algorithms, machine learning, not just to analyze medical data, but to derive insights, insights that allow health care to move beyond sort of an average one size fits all approach. Right. The goal is tailoring treatment diagnosis precisely to an individual patient's unique characteristics, their biology, clinical history, maybe even real time data. Right. And Tempus lays out some pretty ambitious capabilities powered by this AI engine. 1:47 They talk about using AI to, accelerate the discovery of novel therapeutic targets. Yeah. That's huge. Huge for the early stages of drug development. Right? 1:56 Helping researchers find new molecules or pathways. Exactly. And the ambition goes beyond just discovery in the lab. You know? They also aim to use AI for for predicting the effectiveness of treatments for individual patients. 2:08 Which is kind of the holy grail for personalized medicine, isn't it? It really is. Moving from this drug works for most people to this drug is most likely to work for you. Yeah. Based on your specific data profile, imagine knowing with, like, higher confidence which therapy to start with. 2:23 Yeah. Potentially saving valuable time, avoiding treatments that won't work. Mhmm. And another big challenge, especially in cancer care, is connecting patients with clinical trials, potentially life saving ones. Tempest states they can use AI to identify these trials more effectively. 2:40 And that's a massive hurdle, isn't it? Finding the right trial for the right patient at the right time. You're sifting through complex eligibility criteria, matching them to a patient's, you know, intricate medical profile genomics. Exactly. And AI is really uniquely positioned to handle that kind of scale and complexity. 2:58 Okay. And they also aim to diagnose multiple diseases earlier. Earlier diagnosis. That's critical across so many conditions. Yeah. 3:05 Cancer, heart issues. For sure. It can dramatically improve outcomes. And AI could potentially detect subtle patterns in patient data, whether it's genomic clinical imaging data, patterns that might signal the early stages of a disease before it's obvious. For a human, it could easily connect all the dots from different places. 3:22 Precisely. So these capabilities, accelerating discovery, predicting effectiveness, identifying trials, diagnosing earlier, they really are the central promise Tempus is built on, aiming to fundamentally transform personalized patient care using AI as the engine. Right. The engine that processes and makes sense of all this complex biomedical information. And the foundation for all of this, the fuel for their AI Yeah. 3:48 It has to be data. Right? Oh, absolutely. The sources highlight a simply vast amount of data involved, which is crucial for training robust AI models, especially, you know, deep learning networks that need enormous data sets to generalize effectively. Assess we're talking. 4:03 They mentioned around 8,000,000 de identified research records powering their scientific discovery efforts. That's a deep pool of patient info for analysis. And overall, they cite over 300 petabytes of data in total. Wow. 300 petabytes. 4:16 Can you put that in perspective? Well, a petabyte is a thousand terabytes. A terabyte is a thousand gigabytes. So So hundreds of thousands of standard hard drives worth of data. Something like that. 4:25 Yeah. It's an enormous scale. And this scale of multimodal data aggregation integrating different types of information is absolutely critical. It's what you need to train sophisticated AI models that can find connections across different biological and clinical domains. Okay. 4:40 That scale is mind boggling. Yeah. Let's, let's dive into the actual engines driving this. The technology components Tempest highlights. They outline what they call their foundational technology and AI pillars. 4:54 These are the core components they say make their precision medicine approach possible. What are these pillars? Right. They break it down into four key pillars. One, Nexte, lens, and ALGOS. 5:06 And it's important to understand these aren't just abstract concepts. They seem to represent distinct technological functions and platforms within their ecosystem that work together. Okay. Let's start with one. Described as an AI enabled assistant for health care providers and researchers. 5:19 What does that mean in practice? Well, the key function highlighted for one is bringing clinical intelligence right into the EHR the electronic health record. Not the EHR. Yeah. Think of the EHR as the central nervous system of a clinic or hospital, where doctors and nurses spend so much time documenting, accessing patient info. 5:37 And the significance of EHR integration here is massive, isn't it? Totally. Clinicians are already overwhelmed with information and tasks within the EHR. Right. If AI insights were buried in some separate portal or need a whole different workflow Adoption is a huge challenge. 5:53 Yeah. Getting these insights directly into the clinician's existing workflow at the point of care, That's critical. Both a technical and a usability hurdle. Absolutely. Making complex information seamlessly available and crucially actionable. 6:07 So what kind of clinical intelligence is likely being integrated via one one based on what else they do? Logically, it would include highly complex but vital information, like genomic insights from their testing platforms, maybe highlighting specific mutations and their known clinical significance right there in the patient's chart. Okay. It would also likely include potential matches for relevant clinical trials based on the patient's disease profile, genetic markers. Makes sense. 6:33 And it might also provide suggestions for care pathways, maybe based on analyzed real world data from similar patients in their database, perhaps flagging deviations from guidelines or highlighting therapies that have shown better outcomes in similar patient groups. So bringing complex stuff like genomic variants or trial eligibility criteria into a usable format in the EHR, maybe a pop up, or an integrated dashboard. Exactly. That's a significant technical undertaking. Needs robust APIs, careful UI design tailored for those clinical workflows. 7:06 Okay. Moving on to NextE. This pillar is described as having the purpose to identify and close gaps in care. How is that different from one? Well, while one seems focused on delivering specific intelligence at the point of care, NextE sounds more focused on a broader analysis, potentially proactive analysis of patient populations or maybe in visual patient journeys over time to see where care might be suboptimal or, you know, missing. 7:30 Gaps in care. Yeah. What could that mean here? Could mean several things. Maybe identifying patients in a clinic's population who haven't had guideline recommended screenings or vaccinations based on age or risk factors. 7:40 Okay. Preventative stuff. Yeah. Or it could mean spotting potential misdiagnoses by analyzing historical data. Perhaps a pattern of seemingly unrelated symptoms that when you view them together through an AI lens, suggest a specific condition that wasn't initially considered. 7:57 Or maybe flagging patients who could benefit from a different treatment path or a trial they didn't know about based on new data. Exactly. And how would AI identify these gaps? It would likely involve analyzing those vast datasets structured EHR data, maybe even unstructured notes, looking for patterns that deviate from optimal care trajectories or establish guidelines. Comparing individuals to guidelines or outcomes from similar patients. 8:23 Right. Comparing individual patient data against clinical guidelines or aggregated outcomes data from similar patients in their huge database to flag discrepancies. Predictive models might be used to flag patients at high risk of a misdiagnosis or a suboptimal outcome, allowing clinicians to intervene proactively. So that requires pretty sophisticated data integration and analytical models, longitudinal tracking, pattern recognition. Definitely. 8:46 Across diverse data types too. Then there's LNS, its function, find, access, and analyze multimodal real world data or RWD. This sounds fundamental. RWD in Tempest's context is absolutely crucial. It really forms the backbone of their AI training and validation efforts. 9:04 And the sources imply it includes a really diverse array of data types, hence multimodal. Not just siloed datasets? No. It includes genomic data from their sequencing platforms, clinical data pulled from EHRs across their network, potentially imaging data through their pixel series offerings, and digital pathology data from scanned tissue slides. And that term multimodal is key. 9:26 Right? Integrating these different dimensions of patient information. Exactly. Combining a patient's genomic profile genes, mutations, their clinical history, diagnoses, treatments, labs, imaging stands, pathology slides, it provides a much more complete, more nuanced picture of a patient's disease and health status than any single data type could offer alone. And finding correlations across these data types is where the AI really adds value. 9:50 For sure. And Lettuce isn't just about collecting the data. It's about finding, accessing, and analyzing it, which points to sophisticated data warehousing, data harmonization, analytical platforms built on top of this massive integrated dataset. So they're not just data custodians? Doesn't sound like it. 10:06 They provide the infrastructure queries, visualization tools, computational environments for researchers, both internal and external partners, to explore and make sense of this complex data landscape. And we see a concrete example of Linus in action with that real world data collaboration they announced with BioNet. Ah, yes. The mRNA vaccine company. Right. 10:26 That partnership specifically leverages Tempest's robust multimodal datasets, their analytical support, and their computational biology expertise to support Bioness r and d oncology pipeline. So that demonstrates how the Leningess pillar translates into practical use for life sciences partners. They're exploring complex RWD to identify targets, biomarkers, patient cohorts, accelerating drug discovery. It's not just about having the data, but providing the tools and expertise to make sense of it for specific research questions. Precisely. 10:57 Okay. Finally, we have LJOS. These are described as algorithmic models connected to their assays to provide additional insight. Right. This tells us that the AI models under this pillar aren't just general purpose algorithms. 11:09 They are specifically tied to the data generated by Tempus' own testing platforms, their assays. Like their next gen sequencing for genomic profiling. Exactly. That their liquid biopsy assays analyzing circulating tumor DNA. CtDNA. 11:24 That's the DNA shed into the bloodstream by tumors. Right? Less invasive than a tissue biopsy, but technically challenging. That's the one. Less invasive, but you're looking for tiny amounts of tumor DNA mixed with healthy DNA. 11:36 So the Algos are trained on the data coming off their specific tests, likely integrated with their RWD for context and outcome data, all to provide what they call additional insight. And what could that additional insight be? Could entail predictions maybe, how a patient might respond to a particular therapy based on their molecular profile from the assay? Or risk stratification for disease progression, likelihood of benefiting from certain interventions, or identifying specific characteristics from the assay data that are highly correlated with observed outcomes in their large large RWD database. So extracting actual intelligence directly from the test results, but augmented by the context from the bigger database. 12:17 Yeah. These algorithms are likely using machine learning techniques trained on combined genomic, clinical, and outcome data to generate predictive models. Okay. Now let's move into how these foundational technologies and AI pillars are actually put into practice across different medical domains. This is where we look at the specific things to try. 12:34 Tempus' offerings and, importantly, for our focus today, examine the described methods and results, especially from their scientific validation efforts mentioned in the sources. This is where the technical details and the evidence meet. Right. Transitioning from the underlying tech to the applications. Let's start with oncology. 12:52 That seems to be a major focus area for Tempus, and it's where the sources give us the most detailed scientific validation. Okay. In oncology, they list several key offerings, genomic profiling, algorithmic tests, which ties back to Algos, digital pathology, connecting to lenses' multimodal RWD, oncology care pathway solutions, likely using NextE and one one, and clinical trial matching. Mhmm. These are the services built on their platform. 13:18 And the sources point to some significant scientific validation published in a top journal, Nature Biotechnology. These studies give concrete examples of their methods and results using their platform components. Let's break those down. Yes. Definitely. 13:31 Those nature biotechnology studies offer a real glimpse into the technical underpinnings and their impact. Okay. The first study focuses on their XT platform. That's their comprehensive genomic profiling assay. The method described involves extensive molecular profiling combined with clinical data. 13:48 And specifically, the technical detail highlighted is the comparison between paired tumor normal plus transcriptome sequencing, RNA sequencing, versus a less comprehensive approach, tumor only DNA panel testing. Okay. Let's unpack that technical comparison because it's important. Standard tumor only DNA sequencing looks for mutations only in the tumor sample. Right. 14:10 But some genetic variants can be inherited their germ line rather than acquired just in the tumor, which are somatic mutations. Paired tumor normal sequencing means they also sequence the patient's normal tissue, like blood, to identify those inherited variants. Okay. So you can tell the difference between inherited risks and tumor specific changes. Exactly. 14:28 Which is crucial for diagnosing potential hereditary cancer syndromes Mhmm. And making sure targeted therapies are directed at mutations actually driving the tumor's growth. And what about the RNA sequencing part, the transcriptome? Adding transcriptome sequencing or RNA sequencing means they also analyze the RNA molecules present in the tumor. This gives you information about gene expression levels, which genes are turned on or off, and it can reveal important alterations like gene fusions, which might be missed by DNA sequencing alone or are sometimes better characterized by looking at the RNA. 15:00 So the method compares this more comprehensive approach, somatic DNA, germline DNA plus RNA, to a simpler one, just a somatic DNA panel. Correct. Comparing a more technically comprehensive molecular profile to a more standard one. And what were the results stated in that study? The results stated is that this more comprehensive approach increased cancer patients' personalized therapeutic opportunities. 15:23 And, crucially, the technical finding reported was that the paired tumor normal plus transcriptome sequencing approach outperformed tumor only DNA panel testing. It identified more targeted therapies and clinical trials for a large proportion of patients in their study cohort. So that's a direct technical comparison of sequencing methods and their impact on finding actionable insights. Yes. Finding mutations or gene fusions that could potentially be targeted by specific drugs or qualify a patient for a relevant clinical trial. 15:53 It validates that getting that more complete molecular picture, including RNA and the germline context, leads to identifying more potential treatment options. So the technical depth of the assay directly translates into better clinical utility. It's not just doing sequencing, but how you do it matters. Precisely. A very significant technical result published in a high impact journal. 16:15 K. What about the second study? The pan cancer organoid platform, t o? Right. The method here involves showcasing a robust pan cancer tumor organoid platform developed from 1,000 patients. 16:27 Organoids? These are essentially three d miniature versions of a patient's tumor grown in the lab from biopsy tissue. Like mini tumors in a dish. Kind of. They're valuable models because they come directly from patient tumors and are thought to mimic the original tumors complex structure and cell mix better than, say, traditional two d cell lines. 16:45 And developing this from over 1,000 patients, that suggests a pretty significant scale and effort to make this high throughput. Absolutely. Across different cancer types too. And how do they analyze these organoids to see how they respond to treatment? You mentioned AI and imaging. 16:58 Exactly. This study highlights a really technically innovative approach, a neural network based, label free, light microscopy based drug assays. Okay. Break that down. Neural network label free, light microscopy. 17:12 So they expose these patient derived organoids to various cancer drugs in the lab in vitro. But instead of using traditional methods that might involve adding fluorescent labels to track cell death or growth, which can be time consuming, potentially mess with the cells Mhmm. They use standard light microscopy, just regular microscope images to capture pictures of the organoids over time as they're exposed to the drugs. Okay. Then they apply AI algorithms, specifically neural networks, a type of deep learning model, to analyze these images without any special labels. 17:43 The AI is trained to interpret subtle morphological changes in the organoids, changes in size, shape, density, texture, other visual features that indicate whether the cells are dying, growing, or staying the same in response to the drugs. Exactly. Just from the standard light images. This allows for high throughput screening because it's automated and doesn't require complex sample prep beyond standard culture. That is technically sophisticated. 18:06 Mhmm. Using AI to interpret raw, unlabeled microscopy images to assess drug response in patient models at scale. The AI has to learn what visual cues correlate with drug sensitivity and the result. The result is that this platform is capable of predicting patient specific heterogeneity in drug responses. Patient specific heterogeneity, meaning differences even within one patient's tumor. 18:31 Potentially, yes. The significance is huge, not just predicting if a tumor type might respond, but predicting how an individual patient's specific tumor represented by the organoid is likely to respond to different drugs outside the body. It could potentially capture the heterogeneity within a single tumor as different parts might respond differently. Which could potentially inform treatment decisions before a patient even gets a drug. That's the potential. 18:55 Helping clinicians select therapies more likely to be effective for that specific patient's unique tumor biology. It's taking functional testing, seeing how live cells respond to drugs, and making it scalable and predictive using AI and image analysis. Fascinating. So they're validating genomic profiling and a functional modeling platform with AI imaging. What about the third study, liquid biopsy? 19:18 Right. This study involved the validation of a liquid biopsy assay with molecular and clinical profiling of circulating tumor DNA, c c d DNA. As we discussed, liquid biopsies analyze that tumor DNA shed into the bloodstream. Less invasive, useful when tissue biopsies are hard or for monitoring, This study validated their specific technical assay for detecting genetic alterations in ctDNA. Okay. 19:42 And what was the key result or insight from validating their liquid biopsy assay in the context of patient care? The significant finding was that actionable variants would have been missed if only solid tumor or liquid biopsy tests were performed in isolation. Ah, so analyzing both the standard tissue biopsy and a liquid biopsy from the same patient found things that neither test alone caught. Correct. They were able to identify clinically relevant genetic alterations that were not detected by either test alone. 20:08 So the technical insight again is about combining data sources, like multimodal RWD, but here it's combining data from different types of molecular assays. Exactly. It suggests a synergistic effect. Each method might capture different aspects of the tumor's molecular profile that the other misses, maybe due to tumor heterogeneity, ctDNA shedding patterns, technical limits. Which leads to their conclusion. 20:32 That liquid biopsies provide value to patients when used in combination with standard tissue genotyping. It emphasizes that for optimal molecular profiling, integrating data from multiple assay types where technically feasible can be critical for finding those actionable targets. Okay. So these three Nature Biotechnology studies really provide technical validation for different key parts of Tempest's platform. The core genomic profiling method paired to nor more normal plus RNA and its impact. 21:03 A novel AI enabled functional modeling approach, organoids plus label free AI imaging for predicting drug response heterogeneity, and the value of integrating data from different assay types, tissue plus liquid biopsy. Right. And these validation studies directly support the algorithmic models we talked about under the LG West pillar. Those models used for therapy prediction, trial matching, identifying care pathways, they're built on the foundation of the data generated and analyzed by these technically rigorous assays and approaches. Makes sense. 21:33 The algios models aren't just theoretical AI. They're built on and validated by the data and insights from these specific methods. The publications provide the evidence base. Exact time. And the scale of their described impact on oncology is also notable. 21:46 They state they're connected to over fifty percent of US oncologists. That's significant reach. And they report identifying over 30,000 patients for potential clinical trial enrollment, which indicates the platform is actively matching patients to research opportunities addressing that key challenge. Mhmm. And beyond oncology, the sources also highlight applications in other areas showing the platform's breadth. 22:08 Let's look at cardiology. Okay. In cardiology, they list offerings like ECG AI, TempestPixel Cardio, imaging again, solutions to identify under diagnosed, undertreated patients, like the gaps in care idea, and cardiology care pathway solutions. And the significant result highlighted here, a key technical achievement is their FDA five ten k clearance or Tempus ECG AF. FDA clearance. 22:33 510 k. That means they showed it's safe and effective compared to an existing device. Essentially, yes. It demonstrates substantial equivalence to a legally marketed predicate device. And this clearance is for an AI based algorithm applied to standard 12 led ECG data. 22:47 Correct. That common noninvasive heart electrical activity test. And the result is that this algorithm identifies patients at increased risk of atrial fibrillation flutter, AF. AFib. That irregular heartbeat that can cause serious problems like stroke. 23:01 Identifying risk is important, especially for undetected or intermittent cases. Very important. And the technical and regulatory significance here is particularly noteworthy. The sources state this is the first FDA clearance for an AF indication in a specific regulatory category, cardiovascular machine learning based notification software. Okay. 23:22 Let's break that down. First, clearance, specific category, machine learning notification software. Right. It's the first time the FDA cleared software in the specific category that uses machine learning on cardiovascular data, like ECGs, specifically to notify clinicians or patients about an increased risk of AFib. So it's not necessarily diagnosing AFib at that moment? 23:41 Likely not marketed that way. It's more a tool to flag patients who need further investigation for AFib. The technical piece involves ML models analyzing patterns across the whole ECG waveform, possibly subtle variations predictive of future AFib risk even if the ECG itself looks normal at that moment. Different from traditional rule based ECG interpretation. Yes. 24:03 And the regulatory category notification software is key. It defines the intended use, dictating the validation needed. Being first in category for AFib sets a precedent. So a clear example of AI on a common data type, ECG, generating a clinical insight, AFib risk notification, and meeting regulatory standards. Exactly. 24:23 Moving on, radiology applications via the TEPAS Pixel series. TEPAS Pixel therapy response evaluation, lung, breast, cardio. Right. AI applications leveraging imaging data, another key part of that multimodal RWD under lens. For therapy response evaluation, AI models could analyze sequential scans like CT or MRI of a tumor. 24:44 Instead of manual measurement. Yeah. AI could automatically segment the tumor, quantify size volume changes precisely, maybe analyze image textures that correlate with treatment response, but are subtle to the eye. More objective standardized assessment. Potentially. 24:56 And for specific organs like lung or breast. What would pixel lung or pixel breast do? AI could assist in identifying suspicious nodules or lesions on scans, quantifying their features, size, shape, density, borders, track changes, maybe even estimate malignancy probability based on analyzing huge datasets of similar images with known outcomes. Okay. And pixel cardio? 25:20 Similar image analysis on cardiac imaging, quantifying heart chamber volumes, wall motion, identifying risk factors visible in the images. And the technical methods are likely. Deep learning for images, convolutional neural networks. Typically, yes. CNNs or other deep learning models designed for image analysis trained on large annotated medical image datasets to do things like object detection, segmentation, classification within the images. 25:46 So using AI to extract more info from medical scans faster, maybe more sensitive, linking back to that multimodal data concept, imaging alongside genomics and clinical data. Exactly. Building that fuller picture. The sources also briefly mentioned neurology and psychiatry, Tempus Pure, notetaker, and dermatology. Right. 26:02 Showing the platform's potential breadth. For neuropsych, likely using AI on clinical notes, patient data, maybe biomarkers to aid diagnosis or monitoring. Dermatology could involve image analysis of skin lesions, integrating history. Less technical detail provided there in the sources, but it shows the intent to apply the core platform more widely. Exactly. 26:23 Okay. So we've looked at the core tech, the applications, the methods, and results, especially the validation. Now how does Tempest position itself connecting the ecosystem? They seem like a central hub. Yeah. 26:35 They appear to act as a central hub using their tech and massive multimodal data to serve different players, providers, life sciences companies, and patients, kind of facilitating interactions and accelerating progress with a common platform and shared intelligence. Yeah. Okay. For providers, doctors, hospitals, clinics, how do they say Tempus helps them? Sources say they help physicians make more informed treatment decisions, which ties right back to that clinical intelligence via Tempus One in the EHR, distilling complex data genomics, RWD analysis, predictions into actionable insights in the workflow. 27:08 And they mentioned significant reach. 65% of US academic medical centers connected. Over 50% of US oncologists connected through sequencing, trial matching, research partnerships. That's a substantial footprint, especially in oncology and academic settings. Now for life sciences, companies, pharma, biotech, developing new therapies, how does Tempus assist them? 27:30 They state they assist with better drug development, covering research, finding targets, clinical development, designing trials, finding trials, finding patients, and commercialization, understanding real world use. And how do they enable this? By providing access to and analytical support for their core assets. The RWD platform, LENON s, their algorithmic models, LGOS, for prediction or stratification, and potentially services around clinical trial execution or enrollment by finding eligible patients in their network. So RWD helps identify targets or find trial design, understand markets. 28:02 Right. For research, find targets biomarkers. For development, optimize trial design, stratify patients, find sites. For commercialization, understand real world treatment patterns and outcomes. And they emphasize strong industry connections here too. 28:15 95% of the top 20 pharma oncology companies partner with Tempus, over 200 biopharma partnerships total. That signals the life sciences industry sees real value in leveraging Tempus' data and analytical capabilities. And the BioNTech collaboration is that prime example again? Exactly. Using Tempes' robust multimodal datasets, analytical support, and computational biology expertise for BioNTech's r and d pipeline. 28:41 It suggests a deeper collaboration, not just a data sale, providing analytical capabilities and expertise to help a partner solve research challenges. Okay. Finally, patients. How does Tempus describe its connection with them? How do patients benefit directly? 28:56 Sources say they help patients find their own unique and optimal therapy options, including access to clinical trials. Likely involves translating complex molecular and clinical info analyzed by their platform into understandable insights for patient physician discussions and facilitating that crucial trial matching. And they also recently launched a beta for a patient facing app, Olivia. Right, Olivia. Described as an AI enabled personal health concierge app. 29:21 Its stated purpose is empowering patients and caregivers to holistically organize, manage, and proactively take control of their own health data. So maybe aggregating records, test results, imaging reports into one place. Seems like it. Potentially, using AI to help patients understand their data, track their journey, maybe prep questions for their doctor. Giving patients more agency with their complex health info, especially important for complex diseases like cancer. 29:50 The AI likely helps structure and interpret that data for the patient. Yeah. It's an interesting move extending the AI platform directly to the patient level. So looking ahead, Tempus positions itself at the forefront of AI enabled precision medicine using ambitious language like the future of health care. They certainly do. 30:06 And their upcoming webinar mentioned AI and ML in action, demonstrating real world impact in trial design and patient care on 06/12/2025. That suggests an ongoing push to show tangible results. Right. Moving beyond potential, actively showcasing real world impact from their platform and technologies today. In trial design, patient identification, informing treatment decisions, it's a public statement about demonstrating concrete value. 30:32 Which leads us to a final provocative thought for you, the listener, to consider. Think about this integration of vast multimodal data genomic, clinical, imaging, pathology, liquid biopsy via platforms like Tempus. And then consider applying sophisticated AI analysis across all that delivered directly to clinicians in the EHR, predicting drug response from models, identifying patients for trials. How does this scale of data aggregation and AI analysis fundamentally change the practice of medicine and drug discovery? Yeah. 31:01 What are the profound implications for clinical decision making when AI provides insights right in the workflow, influencing tests, therapies? How does predicting drug response in vitro with AI analyzed organoids change treatment selection? And what does AI's increasing ability to synthesize complex data mean for the future roles of clinicians, researchers, even patients in navigating health info and making decisions? It really makes you think about the accelerating pace of change driven by this convergence of big data and AI in health care. Thank you for listening in. 31:31 Subscribe and follow Colaberry on social media links in the description. And check out our website, www.colaberry.ai backslash podcast for more insights like this.
