AI gene editing platform aims to make CRISPR scalable

For years, the promise of CRISPR gene editing has been defined by isolated, miraculous successes. We know it can cure disease, but the process of designing a safe, effective, and specific therapy remains agonizingly slow, manual, and expensive—a classic trial-and-error bottleneck.

A new biotech startup, Cassidy Bio, is betting $8 million that artificial intelligence can finally solve the scaling problem. The Tel Aviv-based company, which announced its launch and seed funding today, is deploying an AI-driven genomic foundation model designed to replace the siloed, manual design process with holistic, predictive reliability.

The $8 million seed round, led by Ahren Innovation Capital, signals significant institutional confidence in this AI-first approach. The funding includes strategic investment from pharmaceutical giants AstraZeneca and Merck KGaA via AION Lab’s venture seeding track, underscoring the industry’s desperate need for a scalable design solution.

The core challenge in gene editing isn't just cutting DNA; it’s ensuring the cut happens exactly where intended (specificity), that the editing system gets to the right cells (delivery), and that the overall therapeutic effect is maximized (efficacy). Cassidy Bio argues that current methods fail because they treat these components—the guide RNA, the enzyme, and the delivery vehicle—as separate problems.

Cassidy Bio’s platform integrates proprietary wet-lab validation data, population-scale genomic insights, and advanced machine learning. The goal is to predict the optimal combination of guides, enzymes, and delivery modalities for a specific therapeutic context. This moves the process from hoping a design works in the lab to simulating and validating it *in silico* first.

“The promise of genome editing will only be realized when we move beyond isolated successes and build a foundation that can scale,” said Rom Kshuk, CEO of Cassidy Bio, a serial biotech entrepreneur. He emphasizes that clinical confidence must be established at the earliest design stages if gene therapies are ever going to reach millions of patients.

This approach is crucial because while sequencing costs have plummeted and we know thousands of mutations, translating that knowledge into reliable therapies is still limited by the complexity of biological interaction. A guide that works perfectly in one cell line might cause dangerous off-target edits in a patient, a risk the AI gene editing model is specifically designed to mitigate by learning from massive, integrated datasets.