Bindwell

We're building better pesticides using AI. The agrochemical industry is stagnant, it's time for a change.
Backed by Y Combinator, General Catalyst, SV Angel, A.Capital, Character, Paul Graham, and others.

The founders [of Bindwell] will probably do alright. They're smart and have a good idea. Paul Graham, on X

Get in touch Careers

Team

We're Tyler Rose (Wolfram Research) and Navvye Anand (Caltech), a scrappy duo of engineers who met at the Wolfram Summer Research Program in June 2023. We're from China and India, respectively, and both close to farmlands in our countries. United by our passion for tackling global problems, we dropped out of high school / college and started Bindwell to transform the agrochemical industry.

Problem

Pesticides are failing us: their usage has doubled since 2000, even though farmland has decreased. Yet, we still lose 20–40% of crops to pests.

Resistance makes things worse: pests evolve resistance, forcing farmers to use even more pesticides to get the same results. This creates a vicious cycle of increasing resistance and collateral damage.

Innovation is stagnant: since the 2010s, fewer than 40 new active ingredients have been introduced. Most "new" pesticides are just minor tweaks of existing chemicals.

Industry left behind: AI is revolutionizing drug discovery; pesticide discovery is overdue for the same transformation because the underlying biochemistry is similar.

Pesticide safety is a problem: the ideal pesticide kills only pests and nothing else — current solutions poison the entire ecosystem along with their targets.

Solution

We use AI to build better pesticides.

We're building a neural network to replace an entire wet lab. Our wet lab in San Carlos conducts high-throughput assays that generate training data for our models. This approach—scaling computation and data rather than relying on manual design—is the fastest path to developing new pesticides.

Our core technology is Pairwell, a unified model for protein-protein and protein-ligand binding with state-of-the-art generalized uncertainty quantification. Current protein-ligand binding models generalize poorly across protein and chemical space. Without uncertainty quantification, the model confidently predicts in regimes where it has no training data, producing false leads. With calibrated uncertainty estimates, we know precisely when predictions are unreliable. This guides our experimental strategy: we generate assay data in underrepresented regions where uncertainty is high, systematically expanding the model's operating range across the full chemical and protein space relevant to pesticide discovery. We also use Foldwell, our AlphaFold replacement that's 4× faster, for structure prediction of pesticide targets.

We're currently developing our first pesticide. More to come soon…