I am an AI Systems Engineer and Multi-Agent Architect operating at the intersection of high-concurrency Go server components and Python AI/ML infrastructure. Currently architecting production backend components, structured-output pipelines, and optimized LLM routing workflows as a Backend AI Engineer Intern at FlyRank AI and Technical Lead at Builders Collective Africa.

My methodology is backed by a B.Sc. in Computer Science from Ambrose Alli University, deep application training through the Learn2Earn NG Fellowship, and completion of the 19-module Universal AI curriculum developed by MIT Open Learning. I combine low-level systems engineering with modern AI routing paradigms and agile delivery methodologies, holding certified credentials in Scrum Fundamentals (SFC™) and Kanban Essentials (KEC™).

Themis: Autonomous AI Legal Risk Platform

• Core Stack: Python, LangGraph, vLLM, AMD Radeon PRO W7900D (ROCm 7.2.1)
• Architecture: Multi-agent evaluation engine utilizing vLLM speculative decoding with a 1.5B draft model to run high-throughput compliance audits on structured documents. Implemented parallel LangGraph agents to decouple code parsing and static rule checking before running a unified state reduction.
• Repository: jezreal-dev/themis

GemmaRoute: Hardware-Aware LLM Router Engine

• Core Stack: Python, FastAPI, Docker, AMD ROCm, Fireworks AI
• Architecture: Open-source, high-performance LLM orchestration engine designed to optimize inference costs and latency. Built as an intelligent dispatching layer for multi-agent systems, classifying query complexity to route routine requests to local hardware and escalating to cloud models only when required.
• Repository: jezreal-dev/gemmaroute | Frontend Dashboard

SupportShield AI: Zero-Trust Enterprise Support Middleware

• Core Stack: Python, Next.js, FastAPI, Vercel Serverless, Fireworks AI (DeepSeek)
• Architecture: A customer support assistant built for secure IT troubleshooting. Features a dual-pass processing framework called "L2E-Shield" that intercepts user payloads, redacts PII data locally via automated scrubbers, and runs an intent routing loop to block prompt injection attempts before generation.
• Repository: jezreal-dev/supportshield-ai

EduPilot: AI-Powered Lesson Planning Engine [1st Place Hackathon Winner]

• Core Stack: TypeScript, Next.js, Docker, PostgreSQL, GitHub Actions
• Architecture: Full-stack copilot that automatically processes raw curriculum standards into structured lesson blocks, solving administrative infrastructure pain points observed during national educational service. Engineered a complete CI/CD pipeline with automated testing.
• Repository: jezreal-dev/edupilot

CleanPadi: Inclusive Civic Waste Management Engine [1st Place Hackathon Winner]

• Core Stack: TypeScript, HTML/CSS, Vanilla JavaScript, WhatsApp API, USSD Gateway
• Architecture: Multi-interface community waste management platform built for Kwara State that integrates formal municipal waste systems with informal local collectors. Includes a functional client-side JavaScript state-machine to accurately simulate localized cellular USSD menu structures (3477723#).
• Repository: jezreal-dev/CleanPadi

ropa-sci: Cyber-Neon Multiplayer TUI Platform

• Core Stack: Go (Golang), WebSockets, Bubble Tea, Lipgloss, P2P Networking
• Architecture: Concurrent CLI gaming server featuring an interactive terminal user interface. Designed a predictive game-theory AI engine and a thread-safe custom WebSocket persistence layer utilizing RWMutex locks to prevent data races during real-time multi-user states.
• Repository: jezreal-dev/ropa-sci