Sylvia G.

Data Scientist @UCB

Cambridge, MA, US

Built and fine-tuned AI tools including Generative models, predictive models and pipelines for drug discovery, supporting multiple cross-functional research, identifying problems and delivering data science solutions Generated de novo compounds by building GenAI model, combining Transformer with VAE (chemical LLM) to optimize drug-likeness, saving time and budget for selecting and synthesizing drug compounds Built graph neural network model enhanced by Attention, predicting drug properties with 2D input, delivering numerical predictions that reduced validation workload in lab from compounds to 44, saving $19M+ assay cost Engineered data pipeline integrating 15+ clinical data sources for graph-based indexing and API workflows, processing 53M records, reducing development cycle time and achieving 62% detection accuracy versus 38% baseline Collaborated with scientists to understand R&D needs and designed customized solutions, translating complex ML mechanism into actionable dashboards and reports as SOP, reducing screening timelines from months to weeks