Tech and Drugs - Podcast

BenchSci’s CEO advertised for a VP of Marketing. Then he decided not to hire one.What happened next opens a much bigger conversation about AI agents, scientific research, and how companies are built.In this episode of Tech & Drugs, I speak with Liran Belenzon, co-founder and CEO of BenchSci, about two transformations happening together: building an agentic research environment for scientists and redesigning the company behind it around AI.We trace BenchSci’s journey from helping scientists choose antibodies to supporting disease biology research and experimental planning. Liran explains how pharma partnerships and advances in AI changed what the company could build—and why he believes adding AI to existing workflows only takes you so far.Then we get into the difficult questions.What happens to management when everyone can direct a digital workforce? How do you maintain trust through layoffs and constant organisational change? And what should remain human when more work can be automated?We also discuss:• Bringing scientific evidence, models and tools into one research environment.• Why BenchSci reconsidered its VP of Marketing hire.• The leadership habits that need to change when building around AI.• AI costs, scientific publishing, data access and intellectual property.• Why investment in AI drug discovery has yet to deliver the clinical pipeline impact many expected.• Why scientific judgment matters when researchers become orchestrators of AI agents.A candid conversation about scientific ambition, business survival, and the people living through the transition.EXPLORE BENCHSCIhttps://www.benchsci.com/SUBSCRIBE TO TECH & DRUGSYouTube: https://www.youtube.com/@TechandDrugsSpotify: https://open.spotify.com/show/2Y4bCwsd8skxXXMkz8JUJLApple Podcasts: https://podcasts.apple.com/us/podcast/tech-and-drugs-podcast/id1784797151Which part of your organisation would you redesign first around AI—and where would you insist on human leadership? Share your thoughts in the comments.#DrugDiscovery #AgenticAI #TechAndDrugs

Can AI compress years of biomedical research into months? And if it can speed up literature reviews, coding and data analysis, what happens when the next step still depends on an experiment?I’m joined by Stephan Reichl, a biomedical data scientist at CeMM and the Medical University of Vienna, to explore the promise and practical limits of AI in science. Stephan co-authored a study asking how much general-purpose AI could accelerate biomedical research—and where the bottlenecks would move.We discuss smarter experimental design, predictive readouts, AlphaFold, SimulateGPT, the challenges of adopting AI in research teams, and a question that matters more as these systems gain autonomy: who decides what “good science” looks like?GUEST AND RESOURCESStephan’s paper, “What are the limits to biomedical research acceleration through general-purpose AI?”:https://www.nature.com/articles/s41598-025-32583-w“GPT-4 as a biomedical simulator” (SimulateGPT):https://pubmed.ncbi.nlm.nih.gov/38909448/MrBiomics, the bioinformatics workflow project mentioned in the conversation:https://github.com/epigen/MrBiomicsFit For Science, Stephan’s podcast with Rob ter Horst:https://www.youtube.com/@FitForScienceAccelerate Europe:https://accelerate-europe.org/Subscribe to Tech & Drugs for conversations about data, AI and the future of drug R&D:https://www.youtube.com/@TechandDrugsWhat part of scientific research would you trust an AI co-scientist to handle—and where would you insist on a human decision? Tell us in the comments.#AIforScience #DrugDiscovery #TechAndDrugs

Kevin Cramer has spent over two decades in lab informatics. He founded Sapio Sciences, unifying LIMS and electronic lab notebooks into a single R&D platform, and now leads Sigmatic Sciences, building autonomous, agentic AI systems and Scout — an AI scientist built for the wet lab.In this episode of Tech & Drugs, Kevin and Thibault Géoui get into why pharma was great at science but treated software as an afterthought, why "how do I get data OUT of the system" was the question nobody asked, and what a post-ELN world actually looks like. They dig into agentic pipelines and why accuracy compounds across agents, the lab-in-the-loop DMTA cycle, "agentic burnout" when you're the gatekeeper for 300 agents, the difference between LLMs and specialized scientific agents, and why the real bottleneck is trust and change management rather than technology. Kevin closes with the bet he'd stake his career on.A conversation about where in silico discovery is heading — and what has to change for scientists to actually use it.━━━━━━━━━━━━━━━━━━━━━━━━━━━━━CONNECT WITH KEVIN CRAMERLinkedIn: https://www.linkedin.com/in/kevincramerveloxity/Sigmatic Sciences: https://www.sigmaticsciences.com/Sapio Sciences: https://www.sapiosciences.com/━━━━━━━━━━━━━━━━━━━━━━━━━━━━━CONNECT WITH THIBAULT GÉOUILinkedIn: https://www.linkedin.com/in/thibaultgeoui/━━━━━━━━━━━━━━━━━━━━━━━━━━━━━Tech & Drugs explores how AI, data, and technology are reshaping pharma and life sciences R&D — in conversation with the people building it.Subscribe for new episodes.#TechAndDrugs #AIinPharma #DrugDiscovery #AgenticAI #LabInformatics #LifeSciences #AIScientist

In this episode of Tech & Drugs, I sit down with Dr. Stephanie Linker and Dr. Philipp Schnee from Merck KGaA, Darmstadt, Germany to explore how AI is changing protein design, drug discovery, and the role of scientists.We discuss de novo protein design, structure prediction, AlphaFold, scientific agents, AI adoption, and what happens when computational models suggest biological solutions that scientists might never have considered themselves.Stephanie works as a Senior Computational Biochemist in the Digital Innovation team at Merck KGaA, Darmstadt, Germany, focusing on AI-driven protein design and structure prediction. Philipp works in the Data & AI environment at Merck KGaA, Darmstadt, Germany, bringing together scientific expertise and technology to develop end-to-end solutions, including a de novo protein design platform.A central theme of our conversation is the move from predicting biology to increasingly being able to engineer it. Stephanie describes proteins as the “machines of the cell”—and explains how de novo design allows scientists to start designing proteins for specific interactions rather than only discovering what already exists in nature.We also discuss what happens when AI challenges established scientific intuition. Stephanie shares an example of applying new protein-design methods to a target that had previously been considered “undruggable”—and how the results challenged those initial assumptions.Beyond protein design, we explore a much bigger change in scientific work. If AI can access information, write code, generate hypotheses and increasingly orchestrate parts of the research process, where does human scientific value move?For Stephanie and Philipp, curiosity, asking the right questions, connecting disciplines, choosing where to focus, and building trust around AI-generated results become increasingly important.Key themes discussed• AI-driven protein structure prediction and de novo protein design• Moving from predicting proteins to engineering new proteins• AlphaFold, co-folding models and protein language models• Exploring targets previously considered “undruggable”• When AI-generated designs challenge scientific intuition• Balancing exploration and exploitation in protein design• Why transparency and trust matter when scientists use AI• Building an internal protein-design platform for scientists• Making advanced AI tools accessible beyond AI specialists• How scientific agents could change research workflows• Why the scientist of the future may increasingly orchestrate AI agents• Bottom-up versus top-down AI adoption• Why AI in science goes far beyond chatbots and LLMs• Open-source models, frontier models and technological dependency• Choosing the right AI model for the right scientific task• Training scientists and the potential role of AI champions• World models and connecting digital models with physical experiments• What AI-enabled science could look like over the next five yearsGuestsDr. Stephanie LinkerSenior Computational BiochemistMerck KGaA, Darmstadt, GermanyLinkedIn: https://www.linkedin.com/in/stephanie-linker-957ba1188/Dr. Philipp SchneeData & AI / Go GlobalMerck KGaA, Darmstadt, GermanyLinkedIn: https://www.linkedin.com/in/philippschnee/

In this Tech & Drugs episode, I sit down with Mathieu Bourdenx, Group Leader at the UK Dementia Research Institute / UCL, to explore how AI agents and AI co-scientists are beginning to change neuroscience research.We discuss brain aging, Alzheimer’s disease, dementia, systems biology, Kosmos, Open Scientist, CBrain, and what it means to work with AI tools that can search literature, analyze data, generate hypotheses, and support scientific discovery.Mathieu’s research focuses on brain aging and the mechanisms that may increase the risk of dementia. In this conversation, we start with the biology: why Alzheimer’s disease is so difficult, why patient heterogeneity matters, why biomarkers are essential, and why preventing or delaying dementia could have a major impact on public health.We then move into the changing role of data and AI in neuroscience. Mathieu explains why biology is not a simple linear pathway, why modern scientists need to understand data science, and how high-dimensional, multimodal data can help us understand complex biological systems.A major part of the episode focuses on AI co-scientists. Mathieu shares his experience working with Kosmos, an AI system designed to support parts of the scientific process: forming hypotheses, searching literature, analyzing data, and iterating across evidence. He explains how the system generated a hypothesis from single-cell transcriptomic data, how the team tested signals across independent datasets, and why wet-lab validation remains the bottleneck.We also discuss CBrain and Open Scientist, including the importance of open-source tools, trusted research environments, patient data protection, and the future of AI agents that can support dementia research at scale.Finally, we explore what this means for the future of science: AI agents in the daily routine of researchers, scientific claim verification, hallucinations, code review, agent-to-agent critique, faster research output, pressure on scientific publishing, and the idea of “digital brains” for scientists and labs.Key themes discussed:- Alzheimer’s disease, biomarkers, and patient heterogeneity- Why prevention may matter as much as treatment- The limits of N-of-1 longevity experiments- Systems biology and high-dimensional data in neuroscience- Why scientists need data science and AI skills- Kosmos and the rise of AI co-scientists- I-generated hypotheses and experimental validation- CBrain, Open Scientist, and open-source AI for dementia research- Trusted research environments and protected patient data- AI agents in everyday scientific workflows- Verification, hallucinations, and scientific trust- How AI may change lab notebooks, publications, and institutional memoryWhy this matters:AI in science is often discussed in broad and abstract terms. This episode grounds the conversation in the real work of neuroscience: reading papers, analyzing data, testing hypotheses, validating findings, and deciding what is worth pursuing in the lab. For pharma, biotech, AI, data, and R&D leaders, it is a practical look at how agentic AI could reshape scientific work without removing the need for biological expertise, experimental validation, and careful judgment.Guest: Mathieu BourdenxRole: Group LeaderInstitution: UK Dementia Research Institute / UCLLinkedIn: https://www.linkedin.com/in/mathieu-bourdenx-21995546/ Website: https://www.ukdri.ac.uk/team/mathieu-bourdenx#AICoScientist #Neuroscience #DementiaResearch #DrugDiscovery #TechAndDrugs