Submit when the Science is Done
Learn how to make agentic AI trustworthy in clinical trials, with your experts in control, always.
In the first episode of the Verisian Community Podcast, Tomás and Henning, along with industry expert Sunil, discuss how new technology can improve the world of statistical programming. They look at how the Verisian Validation Engine ensures programmers can guarantee the integrity of their analyses. The discussion navigates through the evolution of data analytics cas, the potential of AI in clinical programming, and previews Verisian's forward-looking initiatives in clinical reporting.
Guests: Henning Kuich, CTO & Co-Founder of Verisian; Sunil Gupta, strategic advisor to Verisian, statistical programmer and industry consultant with three decades in the field.
Host: Tomás Sabat Stöfsel, CEO & Co-Founder, Verisian
Topics: Verisian's origins and founding story, a walkthrough of the Validation Engine demo, the company's broader mission and statistical programming hub concept, defining code quality in clinical programming versus general software engineering, realistic near-term use cases for generative AI, the 12-month product roadmap, SAS-to-R transition
The full episode can be found on YouTube, Apple Podcasts or Spotify.
Verisian launched roughly five months before this recording, co-founded by Henning Kuich, a biomedical scientist and software engineer who spent seven years at Bayer in IT, infrastructure, and computational research, and host Tomás, whose prior company built an open-source graph database where Henning was a customer. The episode centers on Verisian's first public demo, the Validation Engine, which analyzes SAS logs from a clinical study — whether a single analysis file or an entire trial — and constructs a full dependency graph covering every line of code, dataset, variable, and derivation. This lets a user pick any point in an analysis and trace everything required to produce it, attach log messages directly to the graph, and see how a message upstream in an analysis can be the root cause of problems that surface further downstream, turning debugging from a manual, program-by-program search into a traceable, visual process.
Henning frames Verisian's mission as challenging the assumption that clinical trial analysis can't be made faster, cheaper, and better all at once, arguing this is only true if you keep the underlying process the same. Verisian's answer is to pair the Validation Engine (the quality control and compliance arm, analogous to QA in software development) with a planned Analysis Engine focused on building and modifying studies faster. A headline feature in development, "edit in confidence," would show a programmer in real time exactly which downstream datasets and outputs are affected by a code change before they run it, aiming to prevent the common failure mode where a fix introduces new bugs elsewhere in a large analysis.
Asked to define code quality, Henning identifies robustness (handling unexpected data gracefully) and transparency (code that's understandable to your future self and to others) as core principles shared with general software engineering. The key difference, he argues, is that data analysis code can fail silently: it runs without throwing an error while quietly producing the wrong result, a risk he considers higher in analytics than in classical software development, where at least a defined start and end to the process constrain the problem space. Sunil adds the regulatory compliance lens: in a team structure with source programmers, QC programmers, and a lead ultimately responsible for the whole set, a traceability tool like this formalizes what organizations already do informally through scan logs, but extends it into structured impact analysis that supports faster, better-documented decisions.
Henning's framework for evaluating AI use cases centers on information density and directionality: going from a detailed, information-dense source (code) to something less dense (a summary, a translation, boilerplate) is realistic; going the other way — generating a complete, specific clinical analysis from a natural-language description of a trial — is not, because you're trying to manufacture high-density, highly specific output from comparatively little input. On that basis, Verisian's initial AI use cases are boilerplate code generation for SAS clinical trial programming, SAS-to-R code translation to support the industry's broader open-source shift, and automated code summarization, letting a programmer quickly re-orient themselves in a pipeline they wrote months earlier. Sunil frames his own AI evaluation criteria more simply: is it actually validated, and is it actually useful, arguing both conditions have to be met before a tool earns his team's trust.
Henning lays out Verisian's near-term product direction as four connected components. The statistical programming hub combines the Validation Engine with the new Analysis Engine, eventually processing historical, previously submitted studies to build organized, reusable code libraries by indication or phase, and using that same code-handling infrastructure to fine-tune company- or indication-specific AI agents. The medical writing hub would draft submission document content directly from code and data analysis rather than from existing documentation, which Henning considers a lower-quality starting point. The submission hub envisions giving regulators direct, structured access to explore a study's validated analysis, effectively another layer of quality assurance packaged alongside a submission. Underneath all of it, a monitoring hub would provide a real-time, always-on dashboard of data quality, code quality, and study status as patients and data actually come in, rather than a report run only at the end of the process. Henning notes the current roadmap is built for SAS, with an R-focused build contingent on community demand, something Sunil had specifically asked about early in Verisian's development.
"What we really want to do is make clinical trial analysis faster, cheaper, and better, and that's that magic triangle that people say you actually can't do." — Henning Kuich
"The greatest danger is that there are no errors thrown because the code works but it's doing something which you didn't intend it to do." — Henning Kuich
"We want to have some type of benefit from them... but we have to be careful to make sure that what it's actually doing is actually validated and actually useful for the user." — Sunil Gupta
"If you start with code and you train and fine-tune on code... I think they can be very efficient. Always with the idea that they make the processes more efficient and don't take over the process of creating the science itself." — Henning Kuich
"I'm really excited about the potential possibilities that are being available to SAS programmers, and the things that we can do to better provide support to the team using these types of tools." — Sunil Gupta
What is Verisian's Validation Engine?
A tool that analyzes SAS logs from a clinical study to build a full traceability graph of every line of code, dataset, variable, and derivation, letting programmers trace any point in an analysis back through everything it depends on and connect upstream log messages to their downstream effects.
How is Verisian trying to make clinical trial analysis faster, cheaper, and better at the same time?
By pairing a quality-control Validation Engine with a planned Analysis Engine for building and modifying studies, including a real-time "edit in confidence" feature that shows the downstream impact of a code change before it's run, rather than optimizing the existing manual process incrementally.
Where does generative AI realistically help in clinical statistical programming, according to Verisian's founders?
In moving from high-information-density code to lower-density output: generating boilerplate SAS code, translating code from SAS to R, and summarizing existing analysis code for faster onboarding. Generating a full clinical analysis directly from a natural-language trial description is considered unrealistic in the near term.
What is Verisian's 12-month product roadmap?
A statistical programming hub (Validation Engine plus a new Analysis Engine), a medical writing hub for drafting submission documents from code and data, a submission hub giving regulators structured access to validated analyses, and a real-time monitoring hub tracking data quality, code quality, and study status as data comes in.
Is Verisian's platform built for SAS or R?
Currently built for SAS, since an estimated 98% of clinical studies have historically been done in SAS, but Henning Kuich says an R-focused build is achievable and will depend on how strongly the community signals demand for it.
The Verisian Community Podcast brings together experts in clinical trials to exchange innovative ideas and best practices central to clinical reporting, submission and review. Aligned with Verisian's mission to accelerate the evaluation and market launch of new medical treatments, each episode features expert insights, with guests ranging from statistical programmers to medical writers, to discuss the challenges and opportunities of the latest software and technology.
You can listen to us on Apple Podcasts or Spotify.
See Verisian in action. Share a few details and we’ll be in touch to schedule your demo.
Your request has been received. A member of the Verisian team will contact you shortly to arrange a demo.
