Shaik Abdus Sattar

I’m a Computer Science engineering student focused on turning machine learning research into working systems—across deep learning, AI safety, and practical full-stack delivery. I’m currently developing clinically-informed deep learning pipelines for radiotherapy QA prediction.

As a Research Assistant, I’m building CNN and hybrid deep learning models to predict pre-treatment Gamma passing rates from dose maps and Treatment Plan Complexity Metrics. I design clinically weighted loss functions and structured pipelines to automate radiotherapy QA, and I’m also working on efficient 3D molecular graph kernels to improve scalable molecular similarity and property prediction.

On the product side, I built Aurova, a privacy-first mental health platform using RAG and LLMs for crisis detection and safe response generation with privacy-preserving personalization. I’ve also engineered Orbit Wars, a hybrid deep learning + RL agent using ModernBERT, PPO, and vectorized self-play, and built HASYv2 with a PyTorch CNN that achieved 1st place and 84% accuracy. I aim to grow where strong engineering meets responsible, data-informed AI.