Maharshi Patel

I am an applied machine learning and computer vision researcher focused on developing end-to-end AI solutions that move research into production. I specialize in building robust training pipelines, curating large datasets, and designing automation workflows that support reliable model deployment.

My work spans hybrid convolution-transformer architectures, anomaly detection, and agentic AI; notable achievements include improving detection accuracy from 83% to 99.8% on large-scale remote-sensing data and creating interpretable visualizations to aid domain experts. I have led dataset curation of 30,000+ samples and engineered forecasting and dynamic-pricing models for operational decision-making.

I collaborate effectively across teams, publish survey and comparative research on stress detection and abusive-content classification, and build multi-agent frameworks for collective reasoning; I seek roles where I can apply deep learning, vision, and LLM integrations to solve real-world problems.