Evolutionary Intelligence and Biologics Discovery with Jeremy Agresti

What It Covers

Jeremy Agresti, founder and CTO of Triplebar Bio, explains how microfluidic droplet screening at picoliter scale enables testing hundreds of millions of biological variants per day, unlocking a horizontal platform business model for biologics discovery across antibody therapeutics, cultivated meat cell lines, and precision fermentation proteins.

Key Questions Answered

• •Evolutionary screening scale: Triplebar's microfluidic platform screens hundreds of millions of genetic variants daily by encapsulating cells in picoliter-to-nanoliter droplets controlled by custom chips. This brute-force coverage — analogous to buying every lottery ticket — eliminates the need to design solutions from first principles, compressing multi-year strain development programs into months with small teams.
• •Platform focus discipline: Triplebar deliberately restricts its organism library to CHO cells for antibody production and Pichia pastoris for food proteins, rather than pursuing organism-agnostic flexibility. This constraint means onboarding a second project in the same organism takes roughly 20% of the time the first project required — Agresti's concrete threshold for classifying something as a true platform.
• •Horizontal business model prerequisite: Vertically integrated synthetic biology companies (discovery through consumer product) exist because slow, expensive discovery forces them down the supply chain to justify economics. Faster, cheaper discovery unlocks a horizontal supplier model — analogous to Michelin making tires without building cars — where Triplebar licenses or delivers optimized strains to partners who handle manufacturing and commercialization.
• •Customer discovery via deliberate provocation: When exploring antibody markets, Triplebar assumed binding discovery was a solved problem and stated this directly to potential customers. The rebuttal — that finding a binder is easy but predicting therapeutic viability is not — revealed the actual unmet need. Stating a wrong assumption generates more candid correction than open-ended questions, particularly with technically sophisticated scientific customers.
• •AI-biology data gap opportunity: Current AI biology models succeed where large, relevant datasets exist — protein structure prediction being the primary example. Functional biological data remains scarce, limiting AI utility for engineering applications. Triplebar positions its high-throughput functional screening as a generator of the large, labeled datasets needed to make machine learning models genuinely predictive for biological design problems.