Sowmya Kanchan

I’m an Associate Data Scientist focused on building machine learning workflows that move from research to production. I design and optimize deep generative models for de novo molecule generation, improving efficiency for drug-candidate discovery.

I also strengthen reliability of early-stage screening by developing ADMET prediction models and using constraint-aware generative approaches guided by medicinal chemistry rules. On top of that, I build LLM-powered NLP pipelines that extract chemical properties and bioactivity from scientific literature and patent documents, turning unstructured text into model-ready datasets.

I’ve led work in patent segmentation and SMILES extraction to automate conversion of unstructured patents into usable datasets, and I conduct literature reviews across 50+ papers and patents to drive model improvements. I’m energized by scalable AI applications, especially where NLP and generative modeling can accelerate discovery.