01
Confidential — Q4 2026
Case study in progress
12025
Lumen
Built with
Case study in progress
What I Do
I build AI software, useful websites, and data tools. I focus on making technology easy to use, reliable, and helpful in everyday work.
AI Software EngineeringLLM ApplicationsAI ToolsProduct ThinkingHuman-centered AI
CH. 01Philosophy
Kelvin Fordjour · MMXXVIA Short Statement
On building useful technology without losing the human side.
I turn ideas and research into products people can trust. I take time to understand the problem, then build the simplest tool that can solve it well.
Five Principles
How the work is made.
§ Principles / 01—05
I
Curiosity
I start by asking questions and learning how the problem really works. I read, listen, and think before I write code.
01 / 05
II
Craftsmanship
I care about the small details. I build products that are clear, reliable, easy to maintain, and ready for real users.
02 / 05
III
Experimentation
I test ideas early with small experiments. This helps us learn quickly and avoid spending time on ideas that do not work.
03 / 05
IV
Simplicity
Good products do not need to feel complicated. I remove anything that does not help the user.
04 / 05
V
Useful intelligence
I build tools that help people finish important tasks faster and with less stress.
05 / 05
CH. 02Services
Three ways I can helpOne studio,
three ways to
work together.
I can help with a full project or with one specific part. Whether you are hiring or need help with a project, feel free to get in touch.
I
AI Software Engineering
I build AI apps, search tools, testing tools, and full-stack websites. I can help take an idea from an early prototype to a reliable product.
Tools
Next.js·Django·Python·TypeScript·RAG·GenAI·LLM Apps·Evals
II
Data Analysis
I turn raw data into clear reports, dashboards, and useful answers that help people make better decisions.
Tools
Pandas·Excel·SQL·Statistics·Matplotlib
III
Web Design
I design and build custom websites that look good, work well, and match the person or business behind them. This site is one example.
Tools
Next.js·Tailwind·Design Systems·Framer Motion
CH. 03Projects
06 of 12 shownKelvin's projects,
each with its own
story.
Here are a few projects I have worked on. Each one taught me something new and helped me build better products.
02
Confidential — Q4 2026
Case study in progress
22024
Atlas
Built with
Case study in progress
03
Confidential — Under embargo
Case study in progress
32024
Riven
Built with
Case study in progress
42023
Cipher
Built with
Case study in progress
04
Confidential — Q1 2027
Case study in progress
05
Confidential — Under embargo
Case study in progress
52024
Meridian
Built with
Case study in progress
AMP DIRECT · 2026
Full-Stack · Digital Platform
06
AMP Direct
62026
Full-Stack · Digital Platform
AMP Direct
A platform that helps an agency manage client work and payments.
I designed and deployed a full-stack platform for an agency. It includes login and user roles, client and project management, quotes, invoices, payments, maintenance plans, support tickets, and client dashboards. I also connected Paystack for payments, set up Docker and Coolify deployment, and configured HTTPS for the live app.
Full-stackNext.js frontend and Django API
PaystackPayments, webhooks, and subscriptions
LiveDeployed at ampdirect.online
Built with
Next.js·TypeScript·React·Django·Django REST Framework·PostgreSQL·Docker·Coolify·Paystack·JWT
CH. 04The Process
Six stepsHow an idea
becomes a useful product.
Good work happens in steps. We learn about the problem, make a plan, build a first version, test it, launch it, and keep improving it.
I
Step I
Research
Listen before building
I spend time reading, asking questions, and watching how people work. I learn what the real problem is before I start building.
Field notes · Literature review · Stakeholder maps
II
Step II
Problem framing
Find the real problem
The first request is not always the real need. I work with you to make the problem clear, focused, and possible to solve.
Hypotheses · Success metrics · Non-goals
III
Step III
Architecture
Plan before coding
I map out the main parts of the product, how they connect, what can go wrong, and what it will cost. This gives the team a clear plan before code begins.
Component diagrams · SLAs · Cost models
IV
Step IV
Experimentation
Test ideas early
I test the riskiest parts first with small, low-cost experiments. If an idea does not work, we learn that early instead of finding out after launch.
Eval suites · A/B harnesses · Qualitative review
V
Step V
Deployment
Launch carefully
I release in small steps and watch how the product performs. I track speed, cost, errors, and where users get stuck so we can fix problems quickly.
Canary · Observability · Rollback plans
VI
Step VI
Iteration
Keep improving
A product is not finished when it launches. I use feedback and real usage to make it more useful over time.
Weekly review · Quarterly goals · Annual reset
A Simple Rule
Not every problem needs AI. Sometimes a simple feature or a clear piece of code is the better answer.
CH. 05Now
Current, as of September 2026What I'm
doing now.
A quick look at what I am working on right now. I update this when things change. Want to talk about it? Say hello.
Reading
Designing Data-Intensive Applications — Martin Kleppmann
Reviewing how databases copy and split data.
Building
A RAG evaluation harness in Python
Testing better ways to measure search quality.
Learning
Django + Celery for async task queues
Learning more about backend work for AI apps.
Open to
AI Software Engineer · Data Analyst roles
Remote or based in Accra. Open to longer projects.