I'm an undergraduate Electrical Engineering student at Covenant University.
I lead the Robotics Research Team for Google Developer
Groups on Campus, where we work on
reinforcement learning, vision-language-action models, and low-cost robotic systems. Previously, I
worked with Dr. Daniel Omeiza at the Oxford Robotics Institute on autonomous driving and graph
neural networks.
I'm passionate about teaching and mentorship. I currently teach machine learning to undergraduate
and graduate women of APWEN.
I am the president of the Association of Electrical
and Information Engineering Students and I founded the AEIES student mentorship program to
help students master engineering concepts from first principles.
My vision is to unify machine learning, control theory, and cognitive science to build embodied AI
systems that think, remember, and learn continuously.
I'm currently interested in meta-learning for self-improving policies, and learning-based model
predictive control for mobility and manipulation in legged and aerial robots.
That is, until I understand how to tackle my long-term vision of building truly cognitive systems.
Developed and tested a Fuzzy-PID controller for single- and double-tank liquid level systems,
showing improved performance over conventional PID control in handling nonlinearity, uncertainty,
and multivariable disturbances.
This work compares brute force and gradient descent optimization in neural networks, showing that
brute force achieves better accuracy and lower memory usage, while gradient descent offers faster
convergence—highlighting the potential of hybrid methods for more adaptable and efficient learning.
Developed a hybrid radiological image enhancement technique combining unsharp masking, logarithmic
transformation, and adaptive histogram equalization, which outperforms CLAHE and Wavelet-based
methods in visual quality metrics—offering a more standardized, accurate, and cost-effective
solution for medical imaging systems.
Developed a smart home automation algorithm that combines appliance scheduling with real-time
environmental sensing to optimize electricity use, achieving up to 68% energy savings in
simulations—offering a scalable solution for reducing residential energy consumption and promoting
smart grid integration.
A low-cost, 6-DOF 3D-printed robotic arm designed to rival the performance of commercial
manipulators like the WidowX 250 S—achieving similar reach, repeatability, and payload handling at
just 2.5% of the cost. Fully open-source and built for accessible research, education, and
prototyping.
Vision-based manipulation policy using the DeepMind Control Suite, MuJoCo Playground, and JAX to
enable reliable pick-and-place performance in simulation. The system learns to detect and lift boxes
autonomously, laying a foundation for scalable robotic manipulation with minimal supervision.
Simulates a TurtleBot in Gazebo with precise motion control via PID and robust localization using an
Extended Kalman Filter, all built on the ROS 2 ecosystem.
CamAI is an AI-powered camera system that tracks people, recognizes actions in real time, and sends
alerts during emergencies. Whether it's boosting safety in traffic zones, securing infrastructure,
or enhancing agricultural efficiency, CamAI turns passive monitoring into intelligent response.
A U-Net-based deep learning app for segmenting brain tumors (necrotic, edema, enhancing) in 3D MRI
scans. Uses FLAIR and T1CE images, with a FastAPI backend and Streamlit frontend for easy
interaction and visualization.
A deep learning project that uses a U-Net model to segment clothing items from images of people.
Whether it's shirts, pants, or mystery fashion choices, ClothSegNet finds the boundaries so your
model doesn't have to guess where the pants end and the fashion crimes begin.