Nov 13, 2025 · 48m · no-priors
No Priors Ep. 140 | With Benchling Co-Founder and CEO Sajith Wickramasekara
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of No Priors, Benchling CEO Sajith Wickramasekara discusses how structured data infrastructure and artificial intelligence are revolutionizing biotechnology, from automating laboratory research to reshaping the global economics of drug development.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 22.8% of the talking time here. How this is scored →
The hosts as informed peer 4.7 Guest teaching 4.4 Guest disagreement 1.4 The hosts pushing back 1.8
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 26:25 Rejecting the simplistic AI drug discovery narrative
Saji forcefully dismisses the naysayer premise that AI discovery is just typing in a disease to get a drug, emphasizing that drug development is a multi-step compression game.
Hardest push from the hosts ▶ 26:02 Pressing on the lack of approved AI-discovered drugsSarah directly confronts Saji with the critical industry argument that after years of promises, there are still no approved drugs out the end of the AI discovery pipeline.
Biggest teaching moment ▶ 7:35 Detailing Chinese biotech's commercial milestone with CarvyktiSaji educates Sarah and the audience on how Legend Biotech developed a major cancer immunotherapy in China that J&J successfully licensed and scaled in the US.
The host holds their own ▶ 13:37 Citing Dario Amodei and structural pharma inefficienciesSarah demonstrates deep domain context by citing Dario Amodei's essay and explaining how pharma historically overlooked multi-billion dollar blockbusters like GLP-1s and Keytruda.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Mapping the End-to-End Data Complexity of Drug Development | 4 | 5 | 1 | 1 | Sarah sets the context and asks Saji to define the types of data Benchling captures and describe the current macro cycle in biotech. Saji details the 9,909 steps required after finding a molecule and explains the shift from pandemic-era exuberance into a trough of disillusionment. | |
| The Strategic Rise and Efficiency of Chinese Biotechnology | 4 | 6 | 1 | 1 | Sarah inquires about China's role in the biotech landscape and how Western pharma is responding. Saji educates on China's cost and speed advantages, citing Johnson & Johnson's commercial success with Legend Biotech's Carvykti. | |
| Deconstructing the High Failure Rate and Costs of Drug Discovery | 5 | 5 | 1 | 2 | Saji unpacks why drug development costs exceed $2B with massive late-stage failure rates due to artisanal digital workflows. Sarah adds structural analysis regarding portfolio-level incentives versus single-asset startup survivorship. | |
| Returns on Intelligence and Lessons from Serendipitous Blockbusters | 6 | 5 | 1 | 2 | Sarah challenges the efficiency of large pharma by citing how blockbuster assets like GLP-1s and Keytruda were initially overlooked. Saji agrees and references Dario Amodei's thesis on returns to intelligence in science. | |
| Benchling AI: Integrating Predictive Simulations and Research Agents | 3 | 5 | 1 | 1 | Sarah asks Saji to delineate Benchling AI's core capabilities. Saji explains its dual focus on simulation models and deep research agents that recover lost institutional knowledge from historical experiments. | |
| Co-Scientists vs. Copilots: Paths Toward Scientific Autonomy | 6 | 4 | 2 | 3 | Saji compares biotech autonomy to Waymo versus Tesla and mentions radiology copilots. Sarah pushes back with domain nuances around automated experimental decision-making and the challenge of scientist trust in AI outputs. | |
| Enterprise Pharma's Strategic Stance on Proprietary AI Models | 6 | 5 | 3 | 4 | Sarah presses Saji with the skeptic's view that after years of hype, no fully AI-discovered drugs have emerged from pipelines. Saji firmly rejects the simplistic 'disease-in, molecule-out' premise, emphasizing incremental compression across every stage. | |
| Emerging Business Models in AI Biotech and Data Marketplaces | 4 | 5 | 1 | 1 | Sarah asks about evolving business models in AI biotech. Saji explains why pure-play model companies face commoditization and discusses SaaS-style model distribution and prospective marketplaces for negative experimental data. | |
| Revisiting the Tools vs. Assets Debate in Modern Life Sciences | 6 | 5 | 2 | 2 | Sarah challenges the traditional venture orthodoxy that pharma value resides strictly in assets rather than tools. Saji points to tool giants like Thermo Fisher and Danaher, while Sarah draws parallels to venture capital's own reliance on intuition over systems. | |
| Why Frontier Labs Target Biology and the Anthropic Partnership | 5 | 4 | 2 | 3 | Sarah cynically observes that foundation model labs emphasize drug discovery partly because it is universally unassailable PR. Saji details why biology benefits from LLM architectures and outlines Benchling's partnership with Anthropic. | |
| Navigating Critical Strategic Pivots: Going All-In on AI | 4 | 3 | 1 | 1 | Sarah asks about the internal pivot where Saji's co-founder gave up his management responsibilities to lead Benchling's AI initiatives during a market bust. Saji explains using founder moral authority to drive controversial strategic bets. | |
| Vertical SaaS Playbook: Customer Proximity and Pattern Recognition | 4 | 4 | 1 | 1 | Sarah brings up an observation from a mutual colleague regarding Saji's operational focus. Saji explains his vertical SaaS methodology of partnering deeply with five to ten representative customers to identify universal industry patterns. | |
| Bridging Cultural Divides Between Academic Science and Software | 5 | 4 | 1 | 2 | Sarah asks how to manage the cultural friction between software engineers and academic scientists, sharing an interview question about attitudes toward capitalism. Saji shares lessons on aligning academic incentive structures with commercial enterprise goals. | |
| Cross-Pollinating Tech Storytelling and Pharma Rigor | 5 | 4 | 2 | 2 | Saji argues that biopharma must adopt tech's storytelling approach to humanize its achievements, while tech must learn biopharma's rigor and regulatory safety culture. Sarah shares a conversation where a top AI researcher dismissed high-rigor applications. | |
| Personal Reflections on Agentic Coding and AI Accessibility | 4 | 2 | 1 | 1 | Sarah closes by asking what AI trends Saji finds exciting outside biology. Saji highlights agentic coding tools enabling non-practicing engineers to build whimsically, as well as conversational AI interfaces making tech accessible to older generations. |
Statements from this episode (34)
Benchling serves 1,300 biopharma companies and 7,000 academic institutions
“Today we work with about 1300 biotech and pharma companies scientists at over 7000 academic institutions, universities, all, all around the world. And our software powers, you know, household names like Moderna and Sanofi and Eli Lilly and Regeneron, but also …”
Wickramasekara: Recent years in biotech equal a dot-com level bust
“We're probably like the last couple of years are probably like the equivalent of like the dot com bust happening for biotech.”
Wickramasekara: Novel biotech modalities are in a trough of disillusionment
“I think investors and companies got very excited and put a lot of money into these categories, and we're kind of in the trough of disillusionment for some of them now.”
Wickramasekara: Next decade of biotech will focus on speed and cost
“If the last decade was about sort of biologics and these new modalities, I think the next is going to be about speed and cost.”
Wickramasekara: Major pharma increasingly licenses drug molecules from China
“We've seen this huge uptick in pharma going to China and buying molecules that they typically would have bought from American biotechs.”
Wickramasekara: J&J's cancer drug Carvykti originated from China's Legend Biotech
“Johnson and Johnson partnered with a Chinese biotech called Legend Biotech... They've taken it, and it's a, it's actually a cancer immunotherapy... For multiple myeloma. And like that, that medicine is very commercially successful and widely distributed in the…”
Wickramasekara: Prescription drugs account for 9% of US healthcare spending
“I think nine percent of healthcare spent, prescription drug sales are nine percent of healthcare spending in the U.S.”
Wickramasekara: Clinical trial optimization is a red herring versus molecule quality
“People get very focused on clinical trials because they're like the biggest line item. And they're important, don't get me wrong, but I think it's actually a bit of a red, red herring. Where, ah, yes, there are operational problems. Like, some studies are desi…”
Wickramasekara: Obesity drugs were completely unfundable five years ago
“So GLP-Ins obviously have just transformed obesity as, like, a treatable disease when, by the way, it was, like, a totally unfundable category of things, like, five years ago.”
Wickramasekara: The core science behind GLP-1s has existed since the 1990s
“And so, but the core science for GLP ones was kind of sitting on the shelf in some sense, like it's been known since like the nineties.”
Wickramasekara: GLP-1s will probably be the best-selling drugs of all time
“And then all of a sudden, like we have this category defining medicine that's going to go on to probably be the best selling drug of all, all time.”
Wickramasekara: Benchling released a deep research AI agent over lab data
“And so we've released this deep research agent. It works similar to the deep research agents from Anthropic and other foundation labs. But what it does is it works over Benchling data with the context of the Benchling data model.”
Wickramasekara: Benchling AI saved a customer an 8-month redundant mouse study
“We had a customer that was getting ready to run some mouse studies, and they were looking at 20 different mouse models, and they used a deep research our deep research capability to look at all the historical mouse studies that they had run, and it turned out …”
Wickramasekara: Much of scientific research is lost to institutional folklore
“And so there's so much of science that lives in like folklore and institutional knowledge, and that's kind of lost over time.”
Wickramasekara: AI in scientific discovery will focus on augmentation over autonomy next 1-2 years
“While I would love for that to happen, and I'm maybe more optimistic on a longer time scale, we will get there. I think in the short term, I'm next one to two years, which, you know, already feels like an eternity in AI time. I'm a little bit more bullish on s…”
Wickramasekara: High-stakes AI requires human accountability to bear legal liability
“And truthfully, like at the end of the day, you probably like, you know, with a radiologist, you probably need a human to be accountable For those decisions, it's not just about the technology. Like, someone, someone's gotta be there to, like get sued if somet…”
Wickramasekara: Biology AI has powerful models but lacks a ChatGPT-like interface
“I think we're like, we've got GPT, but there's no chat. Like, that, that's kind of how I think about it. Like, I think the chat, and I mean chat metaphorically, like, that was the interface that made things really take off in, in software. And I don't think it…”
Wickramasekara: Most Biotech Researchers Are Not Yet Using Much AI in R&D
“Most people aren't really using that much AI and R&D yet. They all want to, they're primed to, but there's a lot of concerns about accuracy, IP, security, legal. And I think the farther you go from SF, the like, Larger those concerns, concerns get.”
Wickramasekara: 90% of vertical AI is workflow translation and building trust
“I think in a vertical, I think 90% of the work is actually, like, translation. It's taking something and making science, making sure scientists trust it, it's the right point in their workflow, it's easy to use, and it's accurate.”
Wickramasekara: Large pharma is piloting AI without organizational transformation yet
“Most of the
Large pharma at this point that I've worked with, like, you know, they've got co-pilot and things like that, and they're doing a lot of pilots of different technologies, but I haven't seen their orgs transformed yet.”
Wickramasekara: Large pharma generates model-training data at scale startups cannot match
“They can generate data to train their own models at a scale that most biotech startups can't match.”
Wickramasekara: Large pharma will produce unique proprietary predictive models
“I think while it's early on sort of the agentic how we work side, I think you're going to see very unique models come out of pharma where their computational scientists are building interesting predictive models that, you know, similar to what's happening in t…”
Wickramasekara: Eli Lilly launched TuneLab to share AI models via federated learning
“Eli Lilly put out this announcement about a project they have called TuneLab, where they're taking their internal models and making them available to the broader scientific ecosystem. So a farmer company is saying, hey, you can use our models, but it's give to…”
Wickramasekara: AI drug development requires optimizing every pipeline step, not one-shot generation
“My mental model is like, there are so many steps. Those steps are all cumbersome and difficult. And this is a game of like making each single thing better. And, like, some of the steps matter more than others, like, having the right target or having, like, a g…”
Wickramasekara: Pure AI biology model companies must morph into biopharma
“I think it's unlikely, it's possible, but it's unlikely that sort of, hey, they're just gonna remain pure model companies who just do deals with pharma, where pharma, you know, pays them a hundred million dollars upfront or something like that, and they have f…”
Wickramasekara: Biopharma rarely buys early data due to lack of trust
“You see very, very few transactions of data. That's because no one trusts anyone else's data. You wait till there's a clinical trial, and the data is positive, and you buy the molecule. But you'd think that you'd see a lot more selling of data before that. But…”
Wickramasekara: AI will convert life sciences skeptics to structured data tools
“But now with like AI, one, I think the benefits are a much more immediately obvious to everyone. And so that's going to be this amazing tailwind to try to like do better here. And I think it will convince a lot of people who might have been skeptics in the pas…”
Wickramasekara: Science is uniquely structured to benefit from LLM architectures
“I go back to that sort of returns to intelligence piece where I think science is a problem that has some shape to it that really benefits from the LLM architecture. Like you just think about the corpus of scientific literature as like this vast pool of unstruc…”
Wickramasekara: A co-founder's true power lies in moral authority
“I feel like the power of being a co-founder is actually just in moral authority”
Wickramasekara: 5 to 10 customers represent what an entire vertical industry needs
“I find that having five, 10 customers is actually like a pretty representative model for what the entire industry needs.”
Benchling CEO: Software engineers with bio expertise effectively do not exist
“If I was only hiring like software people who knew bio, I would have like exhausted the pool 10 years ago. It just doesn't exist. We have to take the software people, take the science people, make them sit together.”
Wickramasekara: Biopharma is underappreciated because it communicates as faceless companies
“I think like because sort of the way they communicate is much more about almost these like faceless companies rather than the people, I think it's like easy to hate on them and underappreciate. So I think they need to tell their story and go direct.”
Wickramasekara finds agentic coding tools the most exciting non-bio AI application
“And like, I've tried some of the new agentic coding tools lately, and it's just like, on a personal level, fun to just Feel like the whimsy of being able to build something very quickly again. It's like that. That's pretty cool. That's probably the one I'm mos…”
Guo: Natural Language and Voice AI Accessibility for Older Users Is Undervalued
“I think it's actually undervalued as just because there are audiences that are traditionally not as lucrative as, like, the fast tech-adopting audience of your, you know, 15 to forty-year-old. But it is incredibly wild how the UX of, like, natural language and…”