Wisdom O. Ikezogwo

Leading advanced research in Multimodal Large Language Models (LLMs) for medical imaging and developing cutting-edge generative models.

• Spearheaded research in Multimodal Large Language Models (LLMs) for medical imaging, developing novel medical multimodal datasets (Quilt-1M, MedNarratives) and state-of-the-art analytical models (QuiltNet, Quilt-LLaVA).
• Created a multi-agent AI framework (PathFinder) for clinical diagnosis that outperformed human experts, and established improved benchmarks (MedBlink).
• Directing initiatives to enhance physics-informed image and video generative models, focusing on Newtonian physics for the generation of large-scale video scene graph datasets and pipelines.

Led research on efficient multimodal representations for egocentric data, optimizing data capture costs.

• Spearheaded research into efficient multimodal representations for egocentric data (video, text, audio, IMU, hands), developing the "Perceive-Predict" model.
• Utilized predictive coding between co-occurring modalities to reconstruct missing modalities, significantly reducing data capture costs for expensive modalities like video.

Directed research for a foundational AI model in histopathology, scaling computational efforts and validating models clinically.

• Directed research and development of a foundational AI model for histopathology, leveraging millions of gigapixel histology images.
• Scaled computational efforts on the Argonne National Lab computing cluster and conducted rigorous clinical evaluations to validate model efficacy.

Developed and deployed machine learning models for financial data analysis to enhance lending processes.

• Engineered and deployed machine learning models to analyze unstructured banking data, extracting critical customer earning and spending insights.
• Instrumental in enhancing downstream lending processes, including income prediction, spending pattern analysis, and financial reconciliations.

Designed and implemented production-grade recommendation systems to optimize user engagement.

• Designed and implemented production-grade recommendation systems, leveraging advanced techniques such as attention mechanisms and epsilon-greedy bandit strategies.
• Optimized system performance and user engagement through data-driven model development and deployment.

Conducted research on biosignal processing and neural networks for brain EEG signal analysis.

• Integrated and analyzed disparate multivariate time series data, focusing on characterizing spectral components for dynamic dimensionality reduction.
• Developed and trained neural networks for the classification of brain EEG signals and comprehensive characterization of EEG spectral components.

Designed and implemented a multi-agent AI framework for clinical diagnosis in histopathology.

Developed large-scale medical multimodal datasets (Quilt-1M: 1M Image-Text Pairs, and MedNarratives) to advance research in AI for histopathology.

Developed an efficient multimodal representation model for egocentric data (video, text, audio, IMU, hands).

Led research and development of a foundational AI model for histopathology.

Developed an application for combining disparate EEG seizure datasets into a single, unified dataset.

Awarded $100,000 for pioneering AI research in population health.

Received $20,000 grant to advance research in foundational AI models.

Awarded travel grant to attend the Computational Neuroscience Summer School in Cape Town.

Recognized as the top-performing student in the Control Engineering specialization.

Awarded for outstanding academic achievement as the best male graduating student.

National scholarship awarded for academic excellence (cumulative value: $1500).

Merit-based scholarship awarded for academic achievement (cumulative value: $1500).

Merit-based scholarship awarded for academic achievement (value: $250).

Contributed to research on visual instruction tuning for histopathology videos using localized narratives.

Led the development of a large-scale image-text dataset for histopathology, presented as an oral paper.

Authored an extended abstract on learning compositional representations for histopathology using masked autoencoders.

Contributed to research on risk stratification for solitary fibrous tumors using whole slide image analysis.

Contributed to a collaborative paper reflecting on recent advances and challenges in Machine Learning for Health.

Co-authored research on domain generalization for EEG seizure detection across disparate datasets.

Co-authored research on empirical characterization of temporal dynamics in EEG spectral components.

Forthcoming publication on generating synthetic video scene graphs.

Forthcoming presentation on a novel network for automated triage of colorectal polyps.

Forthcoming comparative study on MsCAMIL performance against pathology trainees in polyp diagnosis.

In submission: A multi-modal, multi-agent framework for diagnostic decision-making in histopathology.

In submission: Research on connecting medical vision and language through narratives across imaging domains.

In submission: Research exploring the fundamental medical imaging knowledge within multimodal language models.

In preparation: Research on efficient multi-modal representations for egocentric data.

In preparation: Research on generating temporally consistent physics in video through alignment.