Computer Science student awarded 2025 Google PhD Fellowship

Rizvi builds foundational AI models for large-scale biological data, with a focus on single-cell genomics. His research addresses a critical challenge in computational biology: bridging the gap between specialized models that process numerical gene expression data and general-purpose Large Language Models (LLMs) that understand biological concepts but cannot directly interpret experimental data. 

Alongside his advisor Prof. David van Dijk, assistant professor of medicine and computer science, and Google’s research team, they have reframed single-cell expression into a natural language format which LLMs can directly ingest. They developed Cell2Sentence, a framework that translates the gene activity profiles of individual cells into a textual format they call "cell sentences," which lists out genes by their name in order of descending expression level. This transformation allows LLMs to process and learn from genomic data as if it were natural language, enabling them to leverage the remarkable reasoning and instruction-following capabilities of modern LLMs, as well as the advanced infrastructure built for training them at a massive scale. As a proof of concept, they recently trained a model on a corpus of over 50 million cell sentences, teaching it to understand cellular biology at an unprecedented scale.

Rizvi’s efforts move us toward a single, versatile AI platform that can analyze, interpret, and predict cellular behavior – and its impact would be significant. The work promises to accelerate drug discovery and therapeutic development by enabling more accurate in-silico simulations of biological systems before conducting physical experiments.