I specialize in engineering structured prompt frameworks and building full-stack AI software applications. By combining computer engineering fundamentals with advanced generative workflows, I turn raw model capabilities into intuitive, user-ready applications.
Holding a B.Eng. in Computer Engineering from Oduduwa University, I specialize in prompt engineering, LLM logic design, and full-stack AI application development. My work centers on translating functional user needs into precise system prompts, agentic workflows, and production software applications.
From AI-moderated social debate software to multi-agent generative visual pipelines and multi-service aggregator apps, I use modern vibe coding, structured frameworks, and rapid prototyping tools to turn abstract ideas into fully operational digital products.
Abuja, Nigeria (WAT / GMT+1) · Global Remote
System Prompts · Agentic Workflows · Context Guardrails · Fallacy Logic
TypeScript · Node.js · Vibe Coding · Supabase · Modern Web Backends
Hardware Diagnostics · Technical Operations · Data Networks · Problem Solving
A React web app for collaborative, one-line-at-a-time storytelling. Writers “Sign the corner” with a unique lowercase username (letters only, no spaces, numbers, or symbols), then leave a single line for someone else to find and continue. The experience is framed like strips of paper passed hand to hand — calm, intentional, and deliberately limited to one line per turn.
Unique author handles are required and can be changed later. The live prototype focuses on lightweight identity and shared chain state rather than full account systems, keeping the barrier to writing as low as possible.
React + Vite, custom paper/ink CSS design system (warm paper backgrounds, seal-red accent, Fraunces / Source Serif 4 typography), and browser storage for local profiles and shared chains. Deployed on GitHub Pages.
Challenge: Most collaborative writing tools become complex dashboards or free-form editors that lose the intimacy of passing a story along.
Solution: Strict one-line turns, a paper metaphor, unique author handles, and an automatic archive lifecycle keep the product focused on the act of writing for the next person — not on managing accounts, permissions, or long-form drafts.
Current live build prioritizes the core loop (sign → write one line → discover on timeline → archive). Richer product directions (AI line scoring, cinematic Watch mode, invite-only Premium chains, full auth) remain on the roadmap and are described in earlier design notes; they are not all active on the public GitHub Pages deployment.
Built with a React / Node.js frontend, using Supabase (PostgreSQL + Auth) as the backend database, rapidly prototyped by generating and refining modular UI components via Claude Artifacts.
Integrated third-party REST APIs into a unified middleware layer to aggregate, parse, and render service data within a single user dashboard. Spans multiple categories — apartments, travel, delivery, dispatch, and professional services — letting users cross-reference options against their own taste and budget in one place. Each result surfaces its actual service provider, and selecting one redirects the user to that provider's site to complete the booking or transaction.
Managed the entire scope from API integration design and data transformation to UI components and serverless deployment.
Challenge: Users had to visit multiple separate websites and apps individually just to find and use the tools or services they needed — apartments, travel, delivery, dispatch, professional services — creating friction and wasted time across a fragmented set of destinations.
Solution: Built a single aggregator app that houses the functionality of multiple third-party services in one unified dashboard, letting users cross-reference options across categories against their own taste and budget to surface the best alternative. Each listing indicates its actual service provider, and clicking through redirects the user to that provider's page to complete the booking — the app handles discovery and comparison, not the transaction itself.
Powered by Anthropic Claude 3.5 Sonnet (selected for complex reasoning and structured rhythmic phrasing in poetry) working alongside OpenAI GPT-4o (configured as a fast, opposing debate persona).
Built using a custom Python agentic workflow using state graphs rather than heavy third-party agent frameworks. The system coordinates turn-taking, dynamic response injection, and context passing sequentially via event-driven state nodes.
React / Next.js frontend with Tailwind CSS for dynamic turn visualization; Python (FastAPI) backend utilizing WebSockets for low-latency streaming of agent outputs.
Formulated the multi-agent orchestration architecture, prompt management system, real-time guardrails, and dynamic UI state sync.
Challenge: Debates — live, upcoming, past, or local — and spoken-word poetry were scattered and hard to discover, leaving audiences with little consistent way to find and engage with them, and limiting how entertaining and far-reaching debate and poetry could actually be.
Solution: Built a platform where users log in, register for a debate by setting the time, topic, and scoring preferences, and can then watch live, upcoming, former, or local debates in one place — turning debate into an accessible, spectator-friendly experience. Extended the same platform to poetry with live poetry challenges, letting poets post and perform live online alongside the debate content.
Designed structural system prompts featuring explicit persona rules and objective constraints (e.g., "Address point B made by Opponent, limit rebuttal to two stanzas, and do not repeat arguments from Turn 1"). This prevented agents from losing topic context or slipping into passive agreement loops.
Powered by Stable Diffusion model APIs for base image generation, combined with a custom Python pipeline using Pillow and OpenCV for image post-processing, trait compositing, and seed variation control.
Python (FastAPI) for the image generation pipeline and batch rendering backend; Node.js / React with Supabase for user management, asset metadata handling, and database storage.
Designed and implemented the system end-to-end, from prompt engineering frameworks and batch rendering scripts to the full-stack user interface and database sync.
Challenge: Maintaining coherent art style, line weight, and feature alignment across high-volume batch generations.
Solution: Used deterministic seed locking coupled with ControlNet conditioning and dynamic negative prompt injection to lock key stylistic traits across randomized asset outputs.
AI Systems Engineer and Prompt Architect with a Computer Engineering foundation — building structured prompt frameworks, LLM workflows, and full-stack AI applications that users can actually operate.
Get a detailed overview of my professional experience, skills, and achievements.
ATS-optimized resume highlighting AI systems engineering, prompt architecture, LLM workflows, and application development experience. Formatted for clear parsing and keyword alignment in Applicant Tracking Systems.
Results-driven AI Prompt Engineer and Application Developer with experience designing multi-stage prompt logic, building full-stack AI web prototypes, and shipping generative workflows into usable products — from AI-moderated debate platforms to multi-character generative art pipelines.
System prompts, agentic workflows, context guardrails, and fallacy-aware reasoning pipelines that keep model behavior reliable in production UIs.
End-to-end AI software with TypeScript, Node.js, Supabase, and modern builders — structured prompts turned into products users can operate.
Computer Engineering background with systems diagnostics, technical operations, and problem-solving across hardware and software environments.
Career path spanning AI application development, technical operations, and computer systems engineering.
Core proficiency across prompt architecture, AI application building, and systems engineering.
Files downloadable
Open to AI Engineering, Prompt Architecture, and AI Application Development roles. Reach out and let's talk.