The work
Five years ago I was writing my first SQL inside a Sierra Leone bank during the Leone redenomination, the kind of project where wrong row counts make front pages. That's still the standard I work against. Today I lead a small team at GTBank Ghana, owning the bank's vendor application stack (ProcessMaker, SWIFT, Alfresco DMS) and shipping in-house automation and data-engineering builds.
My favourite work has a shape: someone is doing something painfully manual, the data exists somewhere, and the right tool would make the work invisible. I rebuilt 65+ business workflows on ProcessMaker (the low-code platform in the Zapier/n8n family) for 500+ users, cutting manual intervention by 30%. I built the DB Change Executor, a Streamlit tool that lets non-engineers safely run schema changes, to remove the engineering team as a bottleneck for routine ops. On the side, I built a self-hosted RAG knowledge-assistant prototype on Llama 3, FastAPI, and Qdrant to learn end-to-end production AI patterns for regulated, on-prem contexts. The same citation-first pattern powers the chat at the top of this site.
How I work
Banking discipline first. Every claim cites its source, every operation runs in a transaction, every system change has a rollback path. I bring the same posture to AI-augmented work. Claude Code is a daily driver, but I write the same kind of tests I would have written by hand, and I read the diff before I trust it.
What I'm curious about
Agentic patterns over LLMs that go past chat. The boring back-office work (document review, exception handling, cross-system reconciliation) is where reliable agents would change the math, and the discipline question is the same one banks already know how to ask: what's the audit trail, what's the failure mode, what's the rollback?
Hybrid retrieval (semantic plus keyword) is underrated. Most "RAG isn't working" stories I read are pure-semantic teams who haven't realised keyword-aware retrieval would solve their entity-recall problem. Policy numbers and acronyms don't embed well.
Outside work
Accra-based. Music is a long-running interest, which is why Riff exists. Football, swimming, and video games for the off-hours.
The thing I most want to get better at outside the job is building real relationships. The kind that turn into long friendships and collaborations, not transactional networking.
Background
BSc Computer Science and Engineering, University of Mines and Technology (Tarkwa). I started writing code that mattered to other people during my Genser internship in 2019, and I've been a working engineer continuously since the Union Systems graduate programme in 2020. Most of what I know I learned on the floor of a bank during cutover.
Reach me
If you're hiring a remote Automation or AI Engineer, or you want to talk shop about RAG in regulated environments, BPM-platform migrations, or letting non-engineers run sensitive ops safely, reach out. Email is fastest: gaddieleacquah@gmail.com. LinkedIn DMs work too: linkedin.com/in/gaddiel-essuman-acquah. The portfolio chat above also takes a swing at most questions about my work.