Bindwell

Market position

Bindwell describes itself as bringing AI-led drug discovery techniques into agrochemical R&D, a sector it characterizes as archaic and slow to adopt AI. It targets incumbents (Bayer, Syngenta, BASF, Corteva are cited as the kind of firms it seeks relationships with) as licensees rather than competitors, and explicitly states it does not compete with vendors of AI models. Y Combinator lists the company as active, based in San Francisco, in the Winter 2025 batch.

Bindwell positions itself as a molecule and IP developer rather than an AI tools vendor, stating it therefore does not compete with companies selling models. Technically, it cites a proprietary model suite spanning the discovery pipeline — Foldwell (structure prediction, reported 4x faster than AlphaFold 3), PLAPT (open-source protein-ligand interaction model able to scan all known synthesized compounds in under six hours), and APPT (protein-protein interaction model for biopesticide screening, reported 1.7x better than existing models on Affinity Benchmark v5.5) — plus an uncertainty quantification system intended to flag when predictions are reliable versus when more data is needed. It contrasts its target-based, computational screening with the industry's practice of synthesizing and testing thousands of compounds empirically.

Technology

Bindwell builds an in-house suite of machine learning models covering the pesticide discovery pipeline, designed with an uncertainty quantification component to reduce unreliable outputs. Foldwell is a custom diffusion-based protein structure prediction model inspired by AlphaFold and reported to run 4x faster than AlphaFold 3, used to identify target structures. PLAPT is an open-source protein-ligand interaction model reported to scan every known synthesized compound in under six hours; it originated as a drug discovery binding-affinity model later cited in a Nature Scientific Reports paper on cancer therapeutics. APPT (Affinity Protein–Protein Transformers) is a protein-protein interaction model for biopesticide screening reported to outperform existing models by 1.7x on Affinity Benchmark v5.5. The company says the models collectively can analyze billions of molecules and return in seconds assay results that would traditionally take days. A wet lab in San Carlos runs high-throughput assays to generate training data and validate predictions, an approach the company frames in terms of scaling compute and data per Richard Sutton's "The Bitter Lesson."

Go-to-market

Bindwell aims to license discovered pesticide molecules and associated IP to established agrochemical companies (named examples of desired contacts include Bayer, Syngenta, BASF, and Corteva). As of November 2025 it was in early discussions with several global agrochemical firms, expected to close a first partnership deal in the near term, and stated a goal of entering licensing deals within about a year. It also recruits engineers for foundational ML model work via its careers page.

Global agrochemical companies that would license Bindwell's discovered molecules and IP (e.g. Bayer, Syngenta, BASF, Corteva are cited as target industry contacts), with farmers as the ultimate beneficiaries; the company has also begun engaging stakeholders in India and China for field testing.

Geography

Headquartered in San Francisco, with a wet lab in San Carlos, California, where high-throughput assays and biological validation of AI-generated compounds are performed. The company has begun discussions with stakeholders in India and China regarding field testing. Its website is offered in English, Chinese, and Hindi.

History

Bindwell traces back to late 2023, when Tyler Rose and Navvye Anand met at the Wolfram Summer Research Program in June 2023 and worked on PLAPT, a binding-affinity prediction model for drug discovery that was later cited in a Nature Scientific Reports paper on cancer therapeutics. In 2024 they began applying the approach to pesticides and founded the company. Bindwell entered Y Combinator's Winter 2025 batch intending to sell AI model access to agrochemical companies, but pivoted to in-house molecule discovery and IP licensing after a roughly 45-minute conversation with Paul Graham at his home. In January 2025 the company published its APPT protein-protein interaction model, and it later launched Foldwell. In November 2025 it announced a $6 million seed round and published a blog post on its first-principles approach to pest control models.

Risks & controversies

Sources note that the company had not found traction with its original model-licensing strategy, as most agrochemical incumbents were reluctant to adopt AI as a core part of pesticide discovery, prompting a pivot. As of November 2025 no licensing deal had closed; the first partnership was only described as expected to close soon, and details of the third-party validation partner were declined. Performance claims (4x faster than AlphaFold 3, 1.7x on Affinity Benchmark v5.5, billions of molecules analyzed) are company-reported. The company is very early stage with a four-person team and its first product candidate still in efficacy testing.