I build multi-agent AI systems, scalable backends, and production-ready applications — from six-agent project monitors to full AI operating systems for engineering workflows.
I'm a software engineer based in Midrand, South Africa. I specialise in building AI-powered applications that combine multi-agent architectures, modern web frontends, and robust backend systems.
My focus is on multi-agent AI systems for engineering and operational workflows — where multiple autonomous agents coordinate to handle real production work. Not single-model wrappers — genuine agentic pipelines.
I've shipped a 6-agent project monitor (PulsePM AI), a full agent OS for engineering workflows (ForgeMind OS), an AI CFO for small businesses (FinPilot AI), an adaptive AI interview simulator (VibeCoach), and a cyberpunk GBA game built in raw C++ (Gridlock 808) — among others.
Evaluating multi-agent AI work for a role or project?
neophukubye2001@gmail.com →Built for the Pinch Hackathon (competing for $50K), FinPilot is an intelligent financial assistant that tells small businesses what their data means and what to do next. 5 coordinated agents handle cash flow prediction, fraud detection, expense analysis, invoice intelligence, and "what if" scenario simulation. Integrates with Pinch Payments for one-click overdue collection.
Built for the AMD Developer Hackathon 2026. Five autonomous agents — Architect, Test Generator, Debugger, Documentation Writer, and Project Planner — coordinate to handle distinct parts of the software development lifecycle. Not a chatbot wrapped around a model; a real multi-agent pipeline where each agent owns a responsibility.
Six coordinating AI agents track project health in real time — surfacing risk before it becomes a blocker and generating executive-ready insights without manual reporting. Built to replace the status-update meeting entirely. Each agent owns a domain: monitoring, risk prediction, timeline analysis, team health, budgeting, and executive briefing.
Not a single coding assistant — a coordinated system of agents handling distinct parts of the software engineering lifecycle, from code generation to workflow execution. The full-scale end of the range: where PulsePM is six agents on one job, ForgeMind OS is the operating layer underneath many. TypeScript-first, modular, and designed to extend. Built during OpenAI Build Week.
Paste a job title and description — VibeCoach generates a full role-specific mock interview: 8 verbal questions + 2 practical tasks, adaptive follow-ups, live webcam PIP, timed phases, and per-question AI scoring with coaching notes. Supports 18 languages. Sessions are persisted with trend analytics, "vs previous session" deltas, and PDF export. Built with Supabase + Gemini.
An AI-powered personal finance assistant that goes beyond budgeting tools — it analyses spending patterns, surfaces savings opportunities, and provides intelligent recommendations on money management. Built in Python with a focus on actionable financial guidance rather than passive data display.
A TypeScript platform that detects misinformation and performs automated fact-checking. TruthLens analyses content credibility, cross-references claims against verified sources, and surfaces confidence scores — built to address the growing challenge of distinguishing reliable information from manipulated content.
An AI-powered intelligent assistant built to provide meaningful, context-aware responses. MaliMind focuses on natural language understanding and dynamic conversation management, built in JavaScript with an emphasis on practical, everyday usability.
An agentic cloud platform that automates Docker containerization and deployment. Instead of writing CI/CD scripts, you declare intent: "Deploy my web service to staging on AWS." The Planning Agent breaks it down, the Execution Engine carries it out — smart code analysis, Dockerfile generation, multi-cloud deployment (AWS ECS, GCP Cloud Run, Azure, K8s), and automatic rollback on failure. Web dashboard included.
An AI-native enterprise command center that fuses DevOps operations with Snowflake Cortex AI — giving teams a unified dashboard to manage infrastructure deployments alongside data pipeline insights, all through a single intelligent interface. Designed for engineering organisations operating at scale.
A full-featured end-to-end and API automated test framework built for production SaaS platforms. Covers UI automation, REST API contract testing, and integration testing — structured to run in CI/CD pipelines and provide reliable regression coverage across complex multi-service architectures.
A high-energy multiplayer cyberpunk arcade game built entirely from scratch in raw C++ for the Game Boy Advance's 25th anniversary (GBA Jam 2026). Think Bomberman meets TRON in a neon terminal grid. 2–4 players connect via GBA Game Link Cable (Multi-Pak mode). Zero AI-generated code, art, or audio — deterministic input syncing via GBA serial registers, hardware sprite palettes, and authentic VRAM layouts.
A hackathon submission for the AMD Radeon Hackathon 2026, built in Python. Leverages AMD Radeon GPU acceleration for high-performance compute workloads — exploring what's possible when AI and GPU-native programming combine in a competitive, time-constrained build sprint.
A web application built for a real bakery business — covering product listings, ordering, and customer-facing functionality. Built in JavaScript with a focus on clean UX, mobile responsiveness, and practical utility for a small business operating in the community.
Open to roles, collaborations, and projects. If you're working on something involving AI systems, backend engineering, or just want to connect — my inbox is open.
neophukubye2001@gmail.com ✉️