Akhil Penumudy

- Fine-tuned and evaluated foundation models from published literature, applying transformer-based architectures to learn representations of protein-protein interaction (PPI) networks and molecular metadata.

- Built scalable RNA-seq QC and single-cell processing pipelines on GPU-accelerated cloud infrastructure, saving 200+ hours of manual effort.

Co-developed a patent-pending ML pipeline for predicting VOC reaction outcomes using chemical graph features and 100K+ molecular sensor interactions.

Optimized downstream purification data collection for GARDASIL9 and built real-time monitoring dashboards across 30+ reactors.

Developed ML-driven biofeedback modules for real-time speech therapy, integrating ultrasound tongue imaging with a Unity game engine.