AI & SOFTWARE FULL-STACK ENGINEER
I operate at the intersection of high-concurrency Go microservices and Python AI/ML infrastructure. Currently architecting intelligent routing engines and robust full-stack applications.
My focus is on driving measurable business value through precision-engineered code, whether that's optimizing LLM inference on AMD ROCm or designing fault-tolerant serverless pipelines.
Executive Summary
Building resilient distributed systems and applied AI products.
I operate at the intersection of high-concurrency Go microservices and Python AI/ML infrastructure. Currently architecting intelligent routing engines and robust full-stack applications.
My focus is on driving measurable business value through precision-engineered code, whether that's optimizing LLM inference on AMD ROCm or designing fault-tolerant serverless pipelines.
REPUTATION & ENDORSEMENTS
"Jezreal is a passionate Go developer who demonstrates remarkable resilience. We interned together, and I can say with certainty that Jezreal is the go-to person for dynamic, high-pressure projects. He consistently delivers exceptional results, even when faced with incredibly tight deadlines."
Micheal Adejoro
Engineering Architect
"Jezreal has always stood out for his analytical mindset and ability to approach complex challenges with confidence. Beyond his technical skills, he is dependable, collaborative, and someone who naturally inspires those around him to aim higher. I highly recommend him to anyone seeking a talented software engineer."
Etsu John
Web Designer & Developer
"Incredible 96 hours! Proud of what we built together ❤️. Appreciate you chasing down that CORS/preflight issue with the middleware—it was brutal but so worth it once the live backend indicator finally lit up."
Nicky Santus
Founder & Lead Engineer, EANO World
"Congratulations on the submission! 👏 Building a local-first AI routing system under hackathon pressure is an impressive achievement. Great work by the whole team."
lablab.ai
AI Hackathons Platform
GEMMAROUTE ROUTING SIMULATOR
Test Jezreal's 4-layer hybrid AI routing engine in real-time. Routine queries execute for free on local AMD ROCm hardware, escalating to cloud models only when required.
Complexity Evaluator
Parses tokens & reasoning depth
Edge Cache & Rules
Instant exact-match lookup
AMD ROCm Gemma 2B
Local Hardware ($0.00)
Cloud Escalation
Fireworks AI / High Reasoning
FLAGSHIP SYSTEM ARCHITECTURES
Explore the end-to-end data flows and component linkages powering Jezreal's AI infrastructure, IoT telematics, and multi-agent platforms.
GemmaRoute Engine
Intelligent 4-layer query router cutting LLM API costs by up to 80% using AMD ROCm local hardware acceleration.
User Query
Frontend / API Client
FastAPI Gateway
Python Router
AMD ROCm Gemma 2B
Local Acceleration
Cloud Escalation
Fireworks AI / Gemini
User Query (Frontend / API Client)
Incoming developer prompt payload.
SHIPPED CODE & LIVE SYSTEMS
GemmaRoute
ProblemHigh API costs and latency bottlenecks when escalating routine developer queries to expensive cloud LLMs.
SolutionAn intelligent 4-layer AI query router that cuts LLM API costs by up to 80%. Routes routine queries through local Gemma 2B running on AMD ROCm for free execution, escalating to cloud models only when quality judges demand higher reasoning.
VoltIQ
ProblemCommercial EV fleets in Lagos lack real-time battery telemetry optimization, leading to grid strain and battery degradation.
SolutionA real-time AI-powered charging optimization engine ingesting high-frequency telemetry via MQTT with LLM agentic charging workflows built on AWS Serverless.
Themis
ProblemManual legal risk assessment and regulatory compliance checks on corporate contracts are slow and error-prone.
SolutionAn autonomous AI legal risk analysis engine utilizing vLLM speculative decoding on AMD Radeon PRO W7900D (ROCm 7.2.1) and parallel LangGraph agents to parse compliance documentation.
RSNA Knee Medical AI
ProblemDetecting subtle structural knee abnormalities from 3D MRI scans requires extensive clinical expertise and multimodal data correlation.
SolutionA multimodal medical computer vision pipeline combining HAIM (Holistic AI in Medicine) clinical embeddings with 3D MRI volume processing in PyTorch to assist diagnostic workflows.
ropa-sci
ProblemTerminal environments lack low-latency, interactive, and predictable multiplayer gaming platforms.
SolutionA cyber-neon terminal user interface (TUI) Rock-Paper-Scissors gaming platform with predictive game-theory AI and thread-safe custom WebSockets.
CleanPadi
ProblemUrban households in Kwara State lack accessible, inclusive waste collection booking channels.
SolutionA multi-channel platform connecting households to verified waste collectors via WhatsApp API and USSD gateways to improve urban sanitation.
TECHNICAL WRITING
Building GemmaRoute: A 4-Layer LLM Router & Firewall
Enterprise AI pipelines are quietly burning capital. When building conversational systems, the default engineering choice is to route everything to massive cloud LLMs. Here is how we cut costs by 80%.
I Built an Autonomous AI Agent That Diagnoses Production Incidents
Five days ago, I started Google's AI Agents Intensive. Today I submitted my capstone, a multi-agent SRE Incident Triage system that autonomously traces production bugs.
The Missing Layer in Modern AI: How Engineers Turn ML Predictions Into Automated Decisions
A predictive machine learning model estimates traffic speeds along three routes, but it is the prescriptive engineering layer that actually makes the decision.
From 2.7 Billion Parameters to Feature Maps: A First-Principles Guide to Computer Vision
Artificial intelligence and computer vision are often presented as recent discoveries, yet their core mathematical principles span decades. A deep dive into CNNs and parameter mapping.
ENGINEERING RADAR
Explore Jezreal's core technologies and architectural paradigms.
Frontend Architecture
Building performant web interfaces with Next.js, TypeScript, & Tailwind CSS. Specializing in global state management (Zustand) and WebSockets for real-time streaming dashboards.
Backend & Cloud
Go & Python for high-concurrency APIs. Extensive experience with Serverless architectures using AWS Lambda, DynamoDB, and Supabase.
Bridging Python's PyTorch & local vLLM/ROCm AI models with high-concurrency Go microservice backends built for extreme throughput and low latency.
AI Systems & Medical Computer Vision
Graduate of the MIT Open Learning “Universal AI” curriculum (3MTT). Building local LLM routers (GemmaRoute), multi-agent risk engines (Themis), and multimodal medical AI pipelines.
Project Management & Leadership
Formally trained in Project Management to effectively lead engineering teams, optimize agile delivery pipelines, and ensure technical roadmaps align strictly with business objectives and timelines.
Verified Codebase Stack
4 ReposHigh-concurrency microservices, goroutines, channels, low-latency I/O
4 ReposAI/ML modeling, local LLM routers, PyTorch computer vision, script pipelines
4 ReposType-safe Next.js 16 dashboards, API integrations, web platforms
1 ReposMedical computer vision, 3D MRI volume processing, HAIM clinical embeddings
1 ReposAWS IoT Core, Lambda, DynamoDB, TimeStream, Amazon Bedrock agentic charging
3 ReposReact 19, Turbopack, App Router, Serverless API routes, web dashboards
2 ReposContainerization, sandboxed testing environments, microservice packaging
2 ReposServer environments, cloud instances, terminal automation, security hardening