Marina Samprovalaki | Agile Actors / Athens University of Economics and Business

About Me

AI Engineer at Agile Actors and Researcher with the AUEB Natural Language Processing Group, specializing in the end-to-end development of Generative AI and Multimodal systems. Expert in architecting Agentic RAG pipelines for complex image and document retrieval, with a deep research background in Multimodal LLMs applied to the biomedical domain. Proven track record in developing high-accuracy Diagnostic Captioning systems. Passionate about bridging the gap between cutting-edge Deep Learning research and scalable, production-ready AI solutions.

Research Interests

Agentic Reasoning: Researching and implementing multi-agent systems using autonomous planning and self-correction loops.
Multi-Agent Orchestration: Investigating decentralized coordination and conflict resolution within agent ecosystems.
Model Context Protocol (MCP): Researching standardized infrastructure for universal interoperability between LLMs and enterprise data tools.
Reinforcement Learning (RLHF/DPO): Investigating Direct Preference Optimization and Reward Modeling for model-human alignment.
Agent-to-Agent Protocols: Investigating standardized communication layers for cross-platform agent collaboration.
Foundation Vision Models: Researching self-supervised pre-training for universal image and video representation.

Thesis

My thesis, titled “Exploring Multimodal Large Language Models for Medical Image Captioning” focused on developing state-of-the-art diagnostic captioning systems for medical images, leveraging the latest advancements in Multimodal LLMs. The research involved designing novel architectures and training strategies to enhance the accuracy and interpretability of generated captions, ultimately aiming to improve clinical decision support and patient care.

The full thesis is available here
The associated codebase here
A slide deck providing a quick overview can be found here

Citation

If you find my work useful, please consider citing it:

@mastersthesis{samprovalaki2024exploring,
  author    = {Samprovalaki, Marina},
  title     = {Exploring Multimodal Large Language Models for Medical Image Captioning},
  school    = {Athens University of Economics and Business},
  year      = {2024},
  pages     = {89},
  url       = {https://nlp.cs.aueb.gr/theses/m_samprovalaki_msc_thesis.pdf}
}