Subhankar Ghosh

Hi there! I'm Subhankar

I am a Computer Science Ph.D. candidate at the University of Minnesota, Twin Cities.

I am a Graduate Assistant at iHARP, where we aim to reduce uncertainty in geospatial forecasting using generative and multimodal AI.

Applied Scientist II Intern, Amazon Science (Summer 2026): Joining the Seattle, WA team to work on generative AI and LLM-based methods for fraud detection. Generative AI Agentic AI LLMs Anomaly Detection

Applied Scientist II Intern, Amazon Science (Summer 2025): Developed generative AI and multimodal fusion methods for anomaly detection in the Amazon Science Geospatial Team (Bellevue, WA). Generative AI Multimodal Fusion Anomaly Detection
Research Intern, Oak Ridge National Laboratory (Spring 2025): Worked with diffusion models and vision transformers for super-resolution tasks. Diffusion Models Vision Transformers
Student Organizer: Scientific Modeling Out-of-Distribution (Scientific-Mood) ML Challenge and Workshop. Scientific ML OOD Generalization
Co-organizer: 2025 NSF HDR Machine Learning Challenge on Anomaly Detection. Pre-print of the paper describing the anomaly detection challenge is out now. Check out our AAAI Workshop on Anomaly Detection in Scientific Domains. Anomaly Detection
Past Research Focus: Investigated spatial pattern mining with statistical guarantees. Pattern Mining Spatial Stats

Spatio-temporal modeling: Forecasting using generative AI Generative AI
Adaptive data integration: Multimodal data fusion strategies Multimodal Fusion
Scalable pattern discovery: Anomaly detection for complex systems Anomaly Detection
Multimodal recommendation: Agents using vision-language models VLMs
Autonomous agents: System design for complex reasoning, planning, and workflow automation Agents
Knowledge grounding: Retrieval-Augmented Generation (RAG) pipelines RAG
Model quantization: Efficiency optimization for foundation models Foundation Models