Adarsh Bhandary Panambur

Turns research models into things people can use — and keeps them running at cluster scale.

• Develop AI solutions built on locally deployed LLMs and VLMs, foundation-model APIs, embeddings and retrieval.
• Build tool-using agent workflows and research software on open-source AI frameworks.
• Engineer training, serving and inference workflows for GPU clusters and supercomputing infrastructure.
• Take AI prototypes from idea to working system quickly, then harden the ones that prove useful.
• Work with academic and industrial partners to translate AI research into practical solutions.

Self-supervised and multimodal learning for mammography and breast MRI, from method design through to publication.

• Developed self-supervised and transfer-learning methods for mammography classification, detection and localization.
• Designed vision-language models for mammography report generation over a curated 81,076-image multi-institutional public dataset.
• Built weakly supervised breast MRI models that localize lesion evidence from a single annotated slice per examination.
• Studied explainability — attention maps, Grad-CAM, token-level grounding — as evidence for clinical review rather than as proof of correctness.
• Published across IJCARS, Scientific Reports, MICCAI, SPIE Medical Imaging, IEEE NSS/MIC and BVM.
• Ran reproducible large-scale experiments with Docker and SLURM on GPU clusters.

Retrospective evaluation of a breast MRI AI research prototype on an external clinical cohort, with radiologists.

• Evaluated a Siemens Healthineers breast MRI AI research prototype on a retrospective external cohort.
• Worked with radiologists on lesion annotation quality and reading-workflow questions.
• Reported the study as an ECR 2025 scientific exhibit (DOI 10.26044/ecr2025/C-21110).

Mammography AI development and retrospective clinical evaluation in international collaborations.

• Contributed to retrospective evaluation of AI-assisted mammography reading in a Vietnamese screening cohort (ECR 2024, ECR 2025).
• Developed deep-learning methods for categorizing mammography positioning artifacts.
• Worked on imaging AI within regulatory and clinical-workflow constraints.

Clinical-workflow prototyping, patent and prior-art analysis, and blockchain-in-healthcare specifications.

• Built MATLAB prototypes for compliance-based task prioritization in clinical workflows.
• Prepared software specifications and use-case survey reports for blockchain applications in healthcare.
• Evaluated IoT technologies for healthcare workflow optimization.
• Identified potential patent infringements, prepared claim charts and conducted prior-art searches.
• Co-inventor on a European patent application for unlocking a medical device (EP 3644550 A1).

Led biomedical engineering delivery and quality assurance services across five Indian states.

• Delivered medical-device QA services across five Indian states.
• Led and mentored a biomedical engineering team.

Hospital equipment testing, preventive maintenance and standards-based compliance work.

• Performed preventive maintenance and compliance testing of medical devices using Fluke biomedical equipment.