TEMPLATE
AI AGENT WORKFLOWS
PROTOTYPE · TYPESCRIPT · MCP · TOOL-USE
A retrieval pipeline and agent orchestration template using MCP tools, function calling, and structured output. Designed for production reuse across multiple products.
▌
MCDARSENEDEV.
Full-stack engineer building
software across the stack — from architecture
and data models to interfaces, infrastructure,
and production AI systems.
// SELF_PORTRAIT.IMG
FULL-STACK // BUILDERv2026.1
SHIPPING
FULL-STACK ENGINEERING•
AI SYSTEMS · AGENTS · RAG•
INFRASTRUCTURE & CLOUD•
PRODUCTION-FIRST MINDSET•
FULL-STACK ENGINEERING•
AI SYSTEMS · AGENTS · RAG•
INFRASTRUCTURE & CLOUD•
PRODUCTION-FIRST MINDSET•
FULL-STACK ENGINEERING•
AI SYSTEMS · AGENTS · RAG•
INFRASTRUCTURE & CLOUD•
PRODUCTION-FIRST MINDSET•
FULL-STACK ENGINEERING•
AI SYSTEMS · AGENTS · RAG•
INFRASTRUCTURE & CLOUD•
PRODUCTION-FIRST MINDSET•
SHIPPED · PRODUCTION
01
MCDARSENE PLATFORM
Full-stack content platform · 2026
▶ PRODUCT_DEMO.mp4
DEMO
STACK
Next.js · SQLite
DEPLOYMENT
Vercel
MODULES
Console · Landing
STATUS
Live
A monorepo content platform with a Next.js admin console, landing page, and shared database package. Console edits trigger toast notifications, persist to GitHub for durability across serverless cold starts, and publish live to the landing page via a shared content API.
SHIPPED · PRODUCTION
02
MVALEX ADMIN
Enterprise admin console · Active development
▶ ADMIN_DEMO.mp4
DEMO
STACK
Next.js · TypeScript
FEATURES
Auth · CRUD · Dashboard
STATUS
In Progress
TYPE
Internal Tool
A production-grade admin console with authentication, role-based access, data management, and analytics dashboards. Built for managing internal operations at Mvalex with a clean, responsive UI and real-time data updates.
01Case Study→
PROTOTYPES · TEMPLATES · UPCOMING
The Lab
EXPERIMENT
REALTIME DASHBOARD ENGINE
PROTOTYPE · REACT · WEBSOCKETS · POSTGRES
A dashboard engine that streams live data through WebSockets with automatic reconnection, optimistic updates, and chart virtualization for 10k+ data points.
UPCOMING
DEPLOYMENT AUTOMATION
TEMPLATE · GO · DOCKER · CI/CD
A deployment automation template that manages infrastructure as code, handles blue-green deploys, and integrates with GitHub Actions for zero-downtime releases.
WHY THE LAB EXISTS
The Lab is where I prototype ideas that don't fit a product scope yet — design experiments, upcoming features, templates, and things that broke in interesting ways. Three of the patterns in production started as a lab experiment. Shipping something usable in a weekend is the best engineering education I know.
DISPATCHES · WEEKLY
Field Notes
/01
Serverless Content Sync: How I Made SQLite Survive Vercel Cold Starts
Vercel's /tmp is ephemeral — every cold start wipes your SQLite database. Here's how I bridged the gap using GitHub as a durable store with the Contents API, syncing content in and out on every request without adding infrastructure.
// sync: GitHub → SQLite on every cold start
await syncFromGitHub();
const content = getPublishedContent();
// ... serve the request
// sync: SQLite → GitHub after every write
await syncToGitHub();VERCELSQLITEGITHUB API
READ MORE→
SEP 30, 202612 MIN READ
/02
From Demo to Dependable: Engineering Decisions That Matter
A working demo and a dependable product are different things. The difference is in the invisible decisions: clear interfaces, observability, security, reproducible builds, and infrastructure costs. Here's how I approach each one.
// the principle: ship what you can maintain
type ActionResult = { ok: boolean; message: string };
// every action returns an honest result
// the UI surfaces it as a toastENGINEERINGARCHITECTUREPRODUCTION
READ MORE→
SEP 15, 20269 MIN READ
// GLOBAL · ENGINEERING SINCE 2018 · STILL BUILDING
OPERATOR PROFILE
One engineer, every layer.
I'm McDarsene M. Mwale — a full-stack engineer who builds software across the stack, from architecture and data models to interfaces, infrastructure, and production systems. My work spans web and mobile applications, backend platforms, cloud infrastructure, and AI systems.
I care about the details that make software hold up in the real world: clear interfaces, thoughtful user experiences, reproducible builds, observability, security, and sensible infrastructure costs. I'm drawn to the difficult, often invisible parts of engineering because they are usually what separates a working demo from a dependable product.
TYPESCRIPTREACTNEXT.JSNODE.JSPYTHONGOPOSTGRESDOCKERAWSAI/MLRAGAGENTS