HELLO THERE 👋

I'm Eyiara Oladipo

Applied ML Engineer | Software Developer

About

Get to know me

Research

View all my research

Projects

See all my applied projects

Contact

Reach out

About Me

oladipoeyiara@gmail.com

H! I'm Eyiara "Ara" Oladipo, a graduate student at the University of Detroit Mercy. I started my journey in computer science with a desire to solve real-world problems. My first major projects included building a faculty reimbursement system used across multiple colleges and creating a course scheduling tool to take the pain out of advising. These experiences fueled my interest in how technology can improve people's lives.

As I delved deeper into AI, I saw how I could integrate it with the tools I was already using, like HTML/CSS, Vue.js, and Node.js. I've worked with TensorFlow, PyTorch, Langchain, and LangGraph to develop end-to-end machine learning models, build data pipelines, and create web applications that make AI more accessible. Throughout my career, I've focused on projects involving natural language processing (NLP), computer vision, and retrieval-augmented generation (RAG) systems. These projects have not only expanded my technical expertise but also deepened my commitment to turning AI research into practical solutions that benefit users, moving beyond theoretical models and ChatGPT wrappers to tangible, real-world applications. While I enjoy research and theory, my true passion lies in developing real-world solutions that help people by solving their problems and simplifying their daily lives.

Research

Centralized and Federated Heart Disease Classification using UCI Dataset: A Benchmark with Interpretability Analysis

Carried out a study evaluating centralized vs. federated learning for heart disease classification using the UCI dataset, evaluating models like SVMs and logistic regression. Incorporated Shapley value-based and LIME-based interpretability to analyze feature contributions, with the goal of furthering privacy-preserving medical ML research.

Python

Scikit-learn

ML Interpretability

Benchmarking Deep Learning Architectures for ECG-Based Multi-label Heart Disease Prediction using MIMIC-IV Database

Applied and optimized CNN, LSTM, and Transformer architectures for heart disease classification on the MIMIC-IV dataset, integrating state-of-the-art methods drawn from 30+ peer-reviewed papers to guide model and training decisions, benchmarking performance, and publishing results at the IEEE CBMS 2025

Python

Pytorch

Tensorflow

CUDA

Joystick-Controlled Convoy Protocol for Secure Communication and Coordination

Developed a joystick-controlled protocol for secure communication and coordination within a convoy of vehicles. The system enables real-time control, enhancing situational awareness and ensuring secure data exchange between vehicles in dynamic environments.

Python

TCP

A Multi-Stage Approach to Lane Detection in Driving Videos Using Classical Image Methods

Implemented a multi-stage approach for lane detection in driving videos using classical image processing techniques. The method combines edge detection, Hough transforms, and thresholding to accurately identify lane boundaries, improving the robustness of lane tracking in varying road conditions.

MATLAB

Signal-Based Access Point Localization Using RSSI-Weighted Clustering and Pattern Similarity

Developed a signal-based method for access point localization by combining RSSI-weighted clustering with pattern similarity techniques. The approach enhances accuracy in pinpointing access points by leveraging signal strength variations and spatial patterns in wireless environments.

Python

Using Machine Learning and Google Earth Engine to Understand Land Use and Land Cover Classifications and NO2 Levels in California

B. Yacoob, E. Scheys, E. Oladipo, A. Price and S. Banitaan, "Using Machine Learning and Google Earth Engine to Understand Land Use and Land Cover Classifications and NO2 Levels in California," 2024 IEEE International Conference on Electro Information Technology (eIT), Eau Claire, WI, USA, 2024, pp. 410-417, doi: 10.1109/eIT60633.2024.10609851.

Python

Google Earth Engine

Navigating Complexity: Hierarchical FSMs for Simulated Robot Control

This research seeks to improve the planning efficiency and scalability of Multi-robot systems (MRS) by expanding on Supervisory Control Theory (SCT) to introduce abstraction and hierarchy.

ROS

MATLAB

Enhancing the University of Detroit Mercy Faculty Reimbursement Management System

Presented at the 2023 University of Detroit Mercy Engineering & Science Research Symposium

HTML/CSS

Vue.js

Node.js

MongoDB

Research Philosophy

My research philosophy centers on creating practical, real-world tools powered by machine learning, tools that don’t just demonstrate theoretical advancements but solve tangible problems for real people. I’m driven by the belief that research should lead to usable, impactful systems, not just papers. For me, success means building solutions that bridge the gap between AI and everyday use.

Applied ML Projects

DaylioAI

Designed and implemented a Retrieval-Augmented Generation (RAG) chatbot leveraging years of my personal journaling data. The system integrates a vector-based semantic search pipeline to retrieve contextually relevant entries, enabling the language model to generate personalized, memory-aware responses. Emphasis was placed on prompt engineering, data preprocessing, and secure local processing with tools like Ollama.

RAG

Langchain

Ollama

Vitain

Helped develop Vitain, an AI-powered platform for personalized supplement recommendations. Leveraging machine learning techniques, the system analyzes user goals, preferences, and health data to suggest tailored supplement plans. By integrating a retrieval-augmented generation (RAG) system with a MongoDB backend, the platform dynamically adapts recommendations based on evolving user inputs and preferences.

Vue.js

Python

RAG

LegalMeasure

Developed LegalMeasure.ai, an platform designed to democratize access to legal resources by simplifying the creation, interpretation, and management of legal documents. The system was built with a focus on usability, explainability, and compliance with jurisdiction-specific legal standards.

Vue.js

Node.js

Stripe