Major AI Project Management platform for a large Architecture Firm.

I've not posted on here for a while. I've been quite busy with a few builds for clients and I cannot believe how it turned out for me personally. I'm super proud of it. Here is one of the builds (I will post the others in the next few days):

This is a full-stack AI-powered operational command center for EPA Architects, a multi national South African architecture firm. This is not a simple project management tool, but a purpose-built, AI-native platform that consolidates every business function of the practice into a single intelligent system. Here is a breakdown of what was built, how it works, and what makes it genuinely AI-powered rather than just AI-branded.

THE TECH STACK

The platform is built on a modern, production-grade stack: React 19 with Tailwind CSS 4 and shadcn/ui on the frontend, Express 4 with tRPC 11 on the backend, MySQL (TiDB) with Drizzle ORM for the database, and an AI Super Agent as the embedded AI brain. Authentication is handled via OAuth with JWT session management. File storage runs on S3. The entire system is deployed.

tRPC provides end-to-end type safety between the frontend and backend — no REST endpoints, no Axios wrappers, no shared contract files. Every procedure is typed from database row to React component. Superjson handles serialization so Date objects and complex types flow through the API without transformation.

FUNCTIONAL MODULES

The platform covers 10 distinct operational areas: Project Management, Time Tracking and Timesheets, Three-Level Visibility Dashboard, Resource Planning and Staff Utilization, Profitability Analytics, Document Management, and ....

THE AI CAPABILITIES - OPERATES AS THE BRAIN

This is where the platform moves beyond conventional project management software. The AI Agent is not a chatbot bolted onto the side of the application. It is embedded throughout as the intelligence layer, operating in two modes: background intelligence and interactive assistance. Call it an AIOS on the platform. It handles Background Intelligence. It processes live project data server-side to generate: project health scores based on multi-variable analysis of budget burn, timeline, and resource allocation; profitability forecasts that project current burn rates forward to completion; utilization insights that identify underutilized staff and over-allocated teams; and portfolio health distributions that give the directors a practice-wide view of which projects are on track, at risk, or in trouble.

It does AI-Powered Planning Recommendations, Context-Sensitive AI Chat Assistant, Automated Data Integration. Unknowingly, I used the "AIOS principle" (now looking at Liam's latest posts).

This build was designed around the AI Operating System (AIOS) framework. The philosophy that AI should not be a set of disconnected tools but an intelligence layer that integrates with the entire organization. I can now say (without initial AIOS planning) that the five AIOS layers are all present in this build:

· Context Layer: All project knowledge, staff data, client information, and organizational structure is captured and structured in the database — giving the AI a rich context to reason from.

· Data Layer: Time entries, attendance records, financial data, and document metadata are all organized and accessible in real time.

· Intelligence Layer: The AI Agent analyses this data continuously, surfacing insights, health scores, forecasts, and recommendations.

· Automation Layer: The various integrations sync, automated timesheet reconciliation, and scheduled report generation remove manual administrative overhead.

Build Layer: The platform itself is the strategic AI asset. It's a custom-built system that gives EPA a competitive operational advantage over firms using generic project management tools.

Most architecture firms manage projects across a combination of spreadsheets, email threads, generic PM tools, and disconnected accounting software. EPA now has a single platform where every project, every staff member, every hour logged, and every rand spent is visible, connected, and intelligently analysed. The directors can see the health of their entire portfolio in one screen. Project managers can see exactly where their budget is going and get AI-powered early warnings before problems escalate. Staff can log time, access documents, and communicate.

0 comments