Exin Therapeutics Is Using AI to Crack the Brain's Hardest Drug Targets

The Macro: The Brain Is the Hardest Organ to Drug

Neurological disorders are responsible for more disability worldwide than any other disease category, and the drugs we have for most of them range from inadequate to nonexistent. Parkinson’s patients still rely on a drug class discovered in the 1960s. Epilepsy associated with autism has limited treatment options. Alzheimer’s drug development has been a graveyard of failed clinical trials for thirty years. The failure rate for CNS (central nervous system) drugs in clinical trials hovers around 90%, which is significantly worse than oncology, cardiovascular, or infectious disease.

The reasons are structural. The blood-brain barrier blocks most molecules from reaching the brain. Neural circuits are extraordinarily complex, and a drug that affects one pathway almost always affects others. Animal models for neurological diseases are notoriously poor predictors of human outcomes. And the diseases themselves are heterogeneous, meaning two patients with the same diagnosis might have fundamentally different underlying biology.

This is why computational approaches to neuro drug discovery have attracted serious attention. Recursion Pharmaceuticals has built one of the largest biological datasets in the world and applies machine learning across therapeutic areas including neuroscience. Insitro, founded by Stanford professor Daphne Koller, combines machine learning with human stem cell models. Verge Genomics specifically targets neurodegenerative diseases with an AI-first approach and has a drug in clinical trials for ALS. Isomorphic Labs, spun out of DeepMind’s AlphaFold work, is applying protein structure prediction to drug design broadly.

But most of these companies are either targeting neurodegeneration specifically (Alzheimer’s, ALS, Parkinson’s) or working across multiple therapeutic areas without deep neurological focus. The intersection of AI drug discovery and neural circuit-level targeting for conditions like epilepsy in autism is still underserved.

The Micro: Oxford Neuroscience Meets Silicon Valley Speed

Exin Therapeutics, from Y Combinator’s Winter 2025 batch, is building an AI drug discovery platform focused specifically on neurotherapeutics. Their approach uses multimodal AI models combined with high-throughput mouse studies to identify drug candidates that target neural circuits rather than individual molecular targets.

The founding team is three Oxford-trained neuroscientists, which is a concentration of domain expertise you rarely see in a YC-backed startup. Gabriel Ocana Santero, founder, holds a PhD in Pharmacology from Oxford with expertise in gene therapy and what the field calls neuroAI. Ivan Lazarte, founder, has a DPhil in Physiology, Anatomy, and Genetics from Oxford with additional training in physics. Marko Tvrdic, co-founder and COO, is also Oxford-trained with a background in developmental and systems neuroscience. Three founders, all from the same university, all with deep neuroscience credentials. That’s either a dream team or an echo chamber, and in biotech, domain depth tends to win.

The company is based in San Francisco with Jared Friedman as their YC partner, which is worth noting because Friedman typically works with technically ambitious companies.

Their platform differentiators, as described on the website, include multi-target effects (hitting multiple relevant pathways simultaneously), highly localized action (affecting specific brain regions rather than the whole organ), genetic background agnosticism (working regardless of patient genetics), and therapeutic modality independence (not locked into one type of drug). Those are ambitious claims, and each one addresses a specific failure mode in traditional neuro drug development.

The website is functional but light on pipeline specifics. There’s a Science section, a Team page, and a Blog, but detailed information about specific drug candidates or development timelines isn’t publicly available. That’s normal for a pre-clinical biotech company. Sharing too much too early in this space invites competitive problems without providing corresponding benefits.

Contact is through [email protected], which is refreshingly direct for a biotech company.

The Verdict

This is the kind of company that either produces something transformative or runs out of money trying. Neuro drug development is expensive, slow, and littered with the remains of well-funded companies that couldn’t translate promising early results into clinical success. YC’s backing helps with speed and connections, but the timelines in biotech are fundamentally different from software.

At 30 days, I’d want to understand the AI platform architecture in more detail. “Multimodal AI for drug discovery” describes half the biotech startups founded since 2023. The differentiator has to be in the specifics of how their models integrate with the mouse study data.

At 60 days, the question is preclinical validation. Has the platform identified candidates that show activity in animal models? Computational predictions without wet lab validation are hypotheses, not drugs.

At 90 days, I’d be watching for partnerships or grants. A three-person team can build the computational platform, but drug development requires resources that extend well beyond what a seed round covers. Strategic partnerships with academic medical centers or pharmaceutical companies would signal that the science is being taken seriously by people who evaluate neuro drug programs for a living.

The strongest thing Exin has going for it is specificity. Rather than building a general-purpose AI drug discovery platform and hoping something sticks, they’ve chosen the hardest organ in the body and said “this is where we’re going to work.” In a field where most failure comes from trying to do too much, that focus is itself a strategy.

I don’t pretend to evaluate the science. But I can evaluate the structure: deep domain expertise, a focused thesis, a hard problem that incumbents haven’t solved, and a technological moment that makes new approaches possible. That combination is worth watching closely.