No PriorsNo Priors Ep. 140 | With Benchling Co-Founder and CEO Sajith Wickramasekara
At a glance
WHAT IT’S REALLY ABOUT
AI agents, biotech data, and rethinking how we develop new drugs
- Benchling CEO Sajith Wickramasekara explains how Benchling became the system of record for biotech R&D, digitizing previously paper-based, bespoke workflows across 1,300+ biotech and pharma companies. He outlines why drug development is so slow, risky, and artisanal, and argues that AI’s biggest near‑term impact will be reducing cost and cycle time rather than magically “discovering drugs” end‑to‑end.
- Wickramasekara details Benchling’s AI strategy: embedding simulation models directly into scientists’ workflows and deploying AI agents over structured R&D data to recommend experiments, surface institutional memory, and automate analysis. He also reflects on macro biotech cycles, China’s emergence as a fast, low‑cost drug engine, and how pharma is cautiously but seriously investing in AI and proprietary models.
- The conversation closes with lessons on building a vertical software company in a complex, regulated domain, integrating scientists and software cultures, and why biotech and pharma must learn to communicate their stories as effectively as leading tech companies.
IDEAS WORTH REMEMBERING
5 ideasDrug development is an extremely long, fragile, and artisanal process.
Going from target identification to commercial drug takes 7–10 years, often costs over $2 billion, and most candidates fail late in clinical trials. Much of the process is bespoke, with every company reinventing workflows and data structures, which drives cost and slows learning.
The biggest near‑term AI gains in bio are about speed and cost, not magic end‑to‑end ‘AI scientists.’
Instead of instantly generating drugs from a text prompt, AI is more realistically going to improve individual steps—better molecule design, smarter experiment selection, faster analysis and manufacturing optimization—compounding into shorter timelines and cheaper programs.
Structured, unified R&D data is a prerequisite for useful AI in biotech.
Benchling’s core value is turning messy, heterogeneous lab data into a consistent, queryable model. Once experiments, molecules, assays, and manufacturing data are standardized, AI agents can surface prior work, answer complex questions, and prevent redundant experiments.
AI agents can unlock institutional memory and prevent wasteful, repeated experiments.
Benchling’s deep research agent, built on models like Anthropic’s Claude, can read years of experimental history and internal reports. In one case, it surfaced prior mouse model studies that would have taken eight months and high cost to replicate, allowing a company to skip redundant work.
Pharma and large biotechs will differentiate via proprietary models and data generation.
Big pharma already runs large‑scale, high‑quality experiments and is starting to train its own predictive models and share them via federated schemes (e.g., Lilly’s TuneLab). Their edge will come from unique internal data combined with targeted, high‑value models, not generic LLM access alone.
WORDS WORTH SAVING
5 quotesIt is probably easier at this point to send things to space or to put people on the moon than it is to get a new medicine approved.
— Sajith Wickramasekara
Right now in bio, we’ve got GPT but there’s no chat.
— Sajith Wickramasekara
Drugs are this amazing ROI… a drug today is only going to get cheaper over time, and it works just as effectively.
— Sajith Wickramasekara
The AI that wins is going to be the one that people actually use.
— Sajith Wickramasekara
If we don’t do this for our customers, who is going to do it?
— Sajith Wickramasekara
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome