01AI / Research
Multimodal
Depression
Detection
A multimodal deep learning system exploring audio and visual data for depression detection through comparative model analysis.
PyTorchTensorFlowOpenFaceVGGish
Let’s Build Something Great
Based in
Nigeria
I care about
Building thoughtful digital products, solving complex problems, and finding the balance between technical quality and a great user experience.
Keep scrolling
Selected projects
I've built real things for real teams across finance, tech, and consulting, shipping frontend and backend work that people actually use.
A consulting / business solutions website presenting services, portfolios, and corporate information in a clean, professional layout.
A financial tech front-end offering banking, payments, loans and wallet services. The site is sleek and responsive, combining interactive components with promotional content to showcase the product offering.
A digital solutions and training company site, showcasing offerings and thought leadership in clean sections and with modern styling.
A static website for a wedding event, holding wedding details, couple details and payment options for event funding
A demo / project site built with React Router (Remix), combining dynamic routing, static generation and advanced Analytics for URL Shortening like tinyurl, bitly, e.t.c.
A full storefront demo showing product listing, detail pages, and eCommerce flows, likely backed by headless APIs and powered by Next.js features.
I am a creative developer and designer focused on building thoughtful digital experiences that combine design, technology, and storytelling.
My work explores the balance between visual expression and functional interaction, creating experiences that feel both intentional and memorable.
Professional journey
Over 4 years of practical experience across consulting, fintech, and software development — building systems, solving deployment problems, and working across the full development lifecycle.
Academic foundation
A foundation in computer science, strengthened through practical development, data systems, and applied machine learning.
A multimodal deep learning system exploring audio and visual data for depression detection through comparative model analysis.
An analytics-driven URL shortener that tracks clicks across location, device, and referrer while preserving user privacy.
A secure, self-destructing data-sharing tool built around client-side encryption and Redis-managed expiry.
Technical toolbox
Languages, frameworks & systems
A collection of technologies I use to build interfaces, backend systems, data pipelines, and machine learning applications.
Scroll to explore
How I work
01 — 05
Start with the problem
Find the right direction
Turn ideas into systems
Break it before users do
Make it real
Impact
01 — 03
PHOS CONSULTS
Reduced repetitive manual follow-up work by helping create more efficient workflows.
PHOS CONSULTS
Improved client engagement through better digital experiences and more effective interaction flows.
EXPERIENCE
Practical experience across frontend development, backend systems, databases, AI, and deployment.
Final year project
Research / 2026
01 — The project
The model uses visual and audio information to identify patterns associated with depression, with a Shifted Window Transformer architecture used as the primary approach.
02 — The model
I compared a Swin Transformer architecture against a CNN baseline to evaluate how effectively the models could distinguish between depressed and non-depressed individuals.
Architecture
Swin Transformer
03 — What mattered
Audio only
Audio features alone performed only slightly better than random guessing.
Visual only
Visual features provided substantially stronger results.
Combining both modalities is what pushed the model performance further.
04 — The result
AUC-ROC — Swin Transformer
Compared with
0.817
CNN baseline
The Swin Transformer outperformed the CNN baseline on 4 out of 5 metrics, including a higher AUC-ROC, showing stronger reliability in distinguishing depressed from non-depressed individuals.
05 — The important part
This is not a clinically validated diagnostic tool.
The next stage is to improve the audio features, strengthen the dataset, and test the system on a Nigerian population before making stronger conclusions about real-world use.
06 — Currently exploring
Outside the projects already in this portfolio, I'm continuously exploring new technologies, better architectures, and ideas that could become the next thing I build.
07 — Technology
The tools change depending on the problem. These are some of the technologies, frameworks, and areas I work with when building applications and experimenting with new ideas.
I don't think the goal is to know every tool. It's knowing how to choose the right ones and learn the rest when the problem demands it.
08 — Contact
Have an idea, a project, or an interesting problem? I'm always open to talking about what we could build together.