Skip to content
Father Godfred Nkayamba

Father Godfred Nkayamba

Systems engineer — payments and AI infrastructure

I design and build systems that have to work every time.

Dar es Salaam, Tanzania·

Open to leadership, architecture and consulting work

About

I started my career more than fourteen years ago writing software for hospital management systems in Tanzania — environments where uptime isn't a metric, it's a matter of care. That discipline of building systems that simply cannot fail is what I bring to every system I work on. What keeps me in this field is financial inclusion: payment systems are how people take part in the economy, so they have to be dependable for everyone — and that is the standard I now bring to building with AI.

What I'm brought in for

Real-Time Payment Systems
Transaction processing that acts as each payment happens, national payment integrations (TIPS, GePG) and mobile money interoperability.
AI Systems & Infrastructure
LLM inference and training pipelines, agent-based systems and AI-native products, built and measured from the kernels up.
Solution Architecture
High-availability distributed systems — microservices, event-driven design, cloud-native and on-prem — designed for failure first.
Technical Leadership & Strategy
Setting technical direction, turning business goals into platform roadmaps and seeing complex systems through from design to production.

Work

Selected work — happy to walk through any of it.

  • SphereCAD editor showing a 3D model of a house, with a command line beneath the viewport

    AI CAD

    SphereCAD

    Open for testing — sign up

    An AI-native 3D CAD platform for small-to-mid buildings, used from the browser. Every change is a command in an append-only log, and a person and an AI planner edit the same model as equal clients.

    Rust · TypeScript · three.js · Python · OpenCascade · PostgreSQL

  • AI infrastructure

    LLM Inference Engine in Rust

    Source on GitHub

    An LLM inference engine written from scratch with no ML framework: its own tensors, SIMD kernels, GGUF loader and quantised model support, behind a CLI and an OpenAI-compatible server. Decode speed is level with llama.cpp on AMD Zen 4 in its own benchmarks.

    Rust · SIMD · GGUF

  • AI agents

    Autonomous Software Company

    A proof of concept of a software company where every role except CEO is an AI agent: agents take an idea through research, planning, build and QA on a single MacBook, with a console for the human approvals.

    Python · FastAPI · LangGraph · PostgreSQL · Next.js

  • SaaS

    Billing & Logistics Platform

    Multi-tenant software for Tanzanian transporters: delivery notes become consolidated invoices, payments and debtor tracking, with TRA fiscal receipts.

    Rust · actix-web · PostgreSQL · Flutter · Typst

Also built: a 343M-parameter language model trained from scratch (see Writing), dbmigrt, a market data feed for the Dar es Salaam Stock Exchange, ClipBeam, Errata and Clocking Tick.

Journey

From hospital systems to national payment rails, and now AI.

  1. 2008–2012

    Foundations

    BSc Computer Engineering & IT · University of Dar es Salaam

    Software development (Java, C) and system design.

  2. 2011–2013

    Systems that can't fail

    Hospital and health research systems

    Built hospital management systems and deployed OpenMRS for HIV/AIDS and TB research programmes.

  3. 2013–2015

    Into payments

    Payment gateways

    Connected banks, mobile money networks and POS behind a single payment gateway. Certificate in National Payment Systems, Bank of Tanzania (2014).

  4. 2015–now

    National scale

    Mobile money platforms

    USSD, real-time triggers, SMS, national payment rails (TIPS and GePG) and a super app. Project Management Professional (PMP), 2026.

  5. 2026

    What's next: AI

    Building my own products

    Now building AI products of my own. SphereCAD is open for testing.

    Try it live

How I work

Correct before clever.

Systems that carry financial transactions must be right first. I design for failure from day one.

Strategy before architecture.

Architecture serves the business. I keep both in view — from the pull request to the boardroom.

Own it end to end.

I stay accountable for what I design — through launch, incidents and every iteration after.

A question I'm thinking about

  • Where does AI genuinely help financial infrastructure — and where should it stay out?

Writing

Articles on architecture and AI, published on LinkedIn.

Architecture

AI

Building payments or AI infrastructure? Let's talk.

Open to technical leadership, solution architecture, and consulting engagements. Based in Dar es Salaam — available globally.

ezrankayamba@gmail.com