Israel Oladeji

Passionate computer science professional with expertise in machine learning and AI. Successfully implemented process improvements at Amazon, enhancing efficiency and mentoring teams. Strong background in deep learning and transformer models, contributing to significant project successes. Adept at leading teams and driving technological innovations.

Motivated and resourceful computer science senior with extensive research experience in narrow AI, deep learning architectures, and transformer-based models, proven leadership, and project management skills through collaborative academic projects, industry internships, and mentorship roles, seeking a challenging role to further develop technical expertise and contribute to cutting-edge innovations while preparing to launch a tech start-up within the next three years

• Successfully transitioned from international education in Nigeria to a rigorous U.S. Academic environment, showcasing adaptability, and determination.
• Consistently engaged in interdisciplinary projects that combine advanced AI research with real-world applications.
• Entrepreneurial Vision: Actively preparing to launch a tech start-up within three years by building industry connections and acquiring comprehensive technical expertise.
• Proven ability to balance academic research, technical projects, and professional work, while maintaining leadership roles and collaborative team efforts.

Deep Learning and Narrow AI Research

• Transformer Enhancements and Decision-Based AI:
• Modified state-of-the-art transformer architectures to improve image recognition accuracy
• Integrated decision-based modules into AI pipelines to enhance inference and adaptability
• Employed frameworks such as TensorFlow, PyTorch, and JAX/Flax to prototype and fine-tune models on standard datasets
• Vision Transformer and MLP-Mixer Architectures
• Contributed to research exploring Vision Transformers (ViT) and all-MLP architectures for image classification tasks
• Investigated data augmentation, regularization, and transfer learning techniques with models pre-trained on ImageNet/ImageNet-21k
• Collaborated with open-source projects to implement fine-tuning pipelines in JAX/Flax, and benchmark performance using tools like Colab

Capstone Project – Mutual Learning for News Classification (CSCI 49500)

• Team Collaboration: Worked with peers (Jorge Puga Hernandez, Avinash Pandey, Brendan Tea) in a structured "teacher-student" framework where multiple models learned collaboratively from distinct data subsets
• Technical implementation:
• Developed a modular codebase (e.g., mutual.py, cli.py) for data preprocessing, training, and evaluation
• Leveraged Python libraries for machine learning to implement mutual learning algorithms, improving classification accuracy on resource-constrained datasets
• Outcomes: Demonstrated enhanced model performance while reducing dependency on large-scale data, validating the mutual learning approach for efficient model training

Additional Technical Projects

• Car Recognition Model:
• Designed and trained a convolutional neural network (CNN) to accurately detect and classify vehicles from images
• Utilized transfer learning and fine-tuning techniques to boost performance on limited datasets
• Weather Prediction System:
• Engineered a deep learning-based system using historical weather data to forecast future conditions
• Incorporated robust data cleaning, feature engineering, and regression analysis to develop a reliable predictive model

• Sunday League soccer player for Abga Ballers