Clara Collier interviews Abhishaike Mahajan, who works at Noetik, a startup focused on developing cancer therapies through AI. Mahajan has a strong background in machine learning applied to healthcare, having spent two and a half years at Anthem, where he worked on predicting chronic health conditions using electronic health records. He then moved to Dyno Therapeutics, where he developed better viral vectors for gene therapy delivery, leveraging models like AlphaFold for protein structure prediction. Mahajan explains the different stages of machine learning in biology: preclinical, clinical, and postclinical. His work at Anthem was postclinical, centered on already approved drugs, while his current role at Noetik involves clinical-stage research on investigational drugs. He emphasizes that the preclinical stage is particularly suited for machine learning due to the lack of established truths, allowing for creative exploration of new molecules or proteins. At Dyno, he focused on improving naturally occurring viruses to enhance their delivery capabilities, specifically targeting areas like the brain. He raises critical questions about scalability and the challenges of working in uncharted scientific territory. The need for models that can generate novel solutions becomes vital, as researchers often deal with evolutionarily divergent viruses and must innovate beyond existing biological frameworks. Mahajan's insights highlight the complex interplay between machine learning and biological research, underscoring the potential and limitations of AI in drug development.