Chief Transformation & AI Officer, eClinical Solutions
The math on clinical trial complexity is unambiguous: data volumes have more than tripled in a decade, driven not by more patients but by more sources—EDCs, central and local labs, wearables, sensors, passive collection from devices patients never bring to a site visit. The problem this creates is not storage. It is comprehension. When data lives in disconnected silos, the burden of synthesis falls on the human beings least positioned to carry it—data managers and medical monitors who should be making decisions, not hunting for information.
Venu Mallarapu, Chief Transformation & AI Officer at eClinical Solutions, has built his argument around that structural gap. elluminate, eClinical’s comprehensive AI-powered clinical data and analytics solution, is his answer to it: a single cloud-based environment that automates data ingestion, standardization, and validation across every source a modern trial produces. In this conversation, Mallarapu discusses what separates high-performing adopters from average ones, why even the ROI case hasn’t accelerated enterprise adoption at large pharma, and what it will mean—for data managers, for regulators, and for patients—when AI agents move from assistants to advisors inside clinical operations.
Moe: Why did the growth in data sources convince you that a unified platform was the right structural answer rather than better point solutions?
Venu Mallarapu: The diversity of data sources kept increasing as trials became more complex. That means you need to source data points from multiple systems and channels, and in order to get to a single source of truth, you need to find a way to aggregate those and then generate the insights required to operate and make decisions.
If that data remains in silos and you are building point solutions, the burden falls on the user—whether it’s a data reviewer, data manager, or medical monitor—to look at data across all of those silos, comprehend it, and then make decisions. That is a laborious and painful process.
That’s where a unified, cloud-based data repository and analytics engine that provides the required intelligence becomes quite helpful. What we were building with elluminate was intended to be that unified clinical data intelligence engine—something that could help manage this complexity and wrangle the data that would be required.
Moe: How does elluminate automate the aggregation of that data across so many different sources and formats?
Venu: That’s kind of the core of our platform. If you drill into the diversity of data sources, there’s the traditional EDC, where studies are set up around the need for sites to work with patients and collect data during visits. But those same patients are also providing samples to labs—local or central—and that lab data is not always coming through the EDC. Then the complexity grew further as more studies wanted to collect data passively through sensors like watches and mobile devices, directly from patients without them needing to come to a site. That became even more of a challenge during COVID.
What we built is a solution that can manage that complexity—you configure what data you’re bringing in, what the structure of it looks like in terms of domains and attributes, and whatever format it arrives in, whether that’s CSV, Excel, or something else. You configure that once, schedule it, and from that point on, whenever data comes in, it’s automatically validated against that standard, any issues are identified, and the data is staged. We’ve also built standards to normalize that data to a specific data model that is conducive to review.
So all of this is automated. You do need to configure things upfront on a study-by-study basis, but once you do that, it runs automatically for the entire life of the study, with only incremental changes as protocols go through amendments. The whole data pipeline is near real-time, or real-time, depending on how frequently you want the data. That’s where it makes life that much easier for data management, data review, medical monitors, and safety teams—they’re not spending time getting the data; they’re spending time on what that data means and what steps they need to take.
“They’re not spending time on getting the data—they’re spending time on assessing what that data means and what steps they need to take.”
Moe: Why do some organizations achieve a 40% reduction in LPLV-to-database-lock time while others achieve closer to 25%?
Venu: We can build the best tool there is, but adoption and how customers use it makes a difference. That becomes even more pronounced when they don’t have the right processes in place. It makes a big difference when customers are re-engineering their processes while bringing in a platform like elluminate—optimizing adoption and use of it.
On the other hand, it also depends on where they’re starting from, because they’re not starting from scratch. They already have an existing process and some level of tooling. Some may be starting at a much lower baseline compared to others, which plays into the percentage improvement they can gain. Somebody who has already optimized their process may see fewer gains compared to somebody who hasn’t optimized as much. Combine that with how well they embed the platform into their processes and manage the change, and that will also impact what that eventual improvement number is for them.
Moe: Why hasn’t enterprise adoption moved faster, given how strong the ROI evidence is?
Venu: The bigger you are, the slower you move—and, I don’t want to throw them under the bus, but the more bureaucratic you can get as well. Scale also limits agility. And you need the right leadership with the vision and strategy to drive the teams. Sometimes it can come down to straightforward budgetary constraints. Some of these are huge transformative programs—you just can’t run those every year, or even every three years.
My feeling is that if an organization believes it has a reasonable solution and a process in place, it can take anywhere between five to ten years for them to decide to pivot to something else and move in that direction. Most of the time they may feel they have things in control and that small incremental changes will be enough. But sometimes they do have to disrupt, and that disruption needs to be supported by leadership vision, strategy, budget, and an honest look at where their organization actually is. It’s a combination of factors.
Moe: How do you expect AI agents to change data management roles over the next three to five years?
Venu: I would say it will affect all of those areas—data mapping, review, RBQM, study ops—and they’re all interconnected and interrelated. You can’t do RBQM without appropriate data mapping and analysis. Audit trail review to identify data fabrication is connected to risk-based quality management. You need to make sure the data is mapped, insights are surfaced, and agents are doing that work. Beyond that, agents will probably identify patterns and anomalies that a human would not otherwise catch.
With regards to how the role will change: we are a heavily regulated industry, and for the right reasons, because everything we do ultimately has an impact on human life—from a safety and efficacy standpoint. For those reasons, I feel that agents will be like assistants, or maybe a step further, advisors, if you will. But the ultimate decision makers are going to be the people who are expected to make those informed decisions, informed by these agentic advisors.
The transition from data manager to data scientist has been talked about quite a bit in the industry. I think this is what makes that a reality. And that would be the case with other functions as well—medical monitors, safety teams—where they will have agents surfacing truths that they might not otherwise be able to get to, while still being in charge of making the final decision. There may be some autonomous agents in certain areas, but all of that is contingent on how you assess the risk associated with that decision. That’s how I think it will evolve over the next two to three years.
“The agents will be like assistants—or maybe a step further, advisors, if you will. But the ultimate decision makers are going to be those people expected to make informed decisions.”
Moe: How does the patient endpoint—faster access to therapy—factor into how you think about platform investment decisions?
Venu: That’s a huge deal. While most, if not all, pharma companies are for-profit organizations and should be thinking about how they benefit their investors, stakeholders, and employees, none of the companies I know of are losing sight of why they exist—which is to ensure that patients are getting the right medicines and treatments at the right time, so that they can live longer, be healthier, and be happier.
Bringing in platforms like this, especially if they tap into the power of AI, will reduce that cycle time. And every day saved is, by some estimates, around ten million dollars in revenue, depending on the therapeutic area—because sponsors have a window of patent exclusivity. So there’s a clear commercial benefit.
But given that it’s the patient we all have in mind to serve, what these agents and AI and unified data intelligence platforms need to ensure is that patient interests are safeguarded—that the data is collected the right way, reviewed the right way, and that appropriate steps are taken to ensure there is no bias involved. And eventually, as newer capabilities like sensors and devices become standard, and as things like omics and genomic sequencing data become easier to collect for specific patients, all of that gets factored in.
When these medicines come to market faster, it benefits sponsors and it definitely benefits patients. It could even make treatments more affordable, because reducing the burden on sponsors reduces the cost of bringing therapies to market, and that would ultimately have an impact on how they price and make treatments available. And from a regulatory standpoint, I think platforms like elluminate could become continuous clinical data intelligence engines for regulators as well—so they’re not waiting on data but receiving it on a continual basis, which would allow them to monitor, oversee, and eventually approve things faster, having been plugged in throughout the entire life cycle of that medicine.
Venu Mallarapu is Chief Transformation & AI Officer at eClinical Solutions, where he leads the company’s enterprise AI strategy.
Moe Alsumidaie, MBA, MSF, is founder and Chief Editor of Vanguard Publications, which publishes Clinical Trial Vanguard, Pharma Vanguard and BullScope, and Head of Research at CliniBiz. He has two decades in clinical trial operations and data science, with earlier roles at Genentech, Abbott Vascular and Stanford University Medical Center, and is a guest lecturer in clinical trial sciences at Rutgers University.
