AI Systems Architect • Macro-Agent Builder
Web Access • Tool Use • Verification • Retro-Action • Systems Engineering
I build connected AI systems that combine web interfaces, backend services, database integration, robotics control, simulation, and agent-style automation.
My current focus is a basic AI macro-agent architecture: an AI model receives a goal, searches or reads context, plans actions, executes safe steps, verifies the result, and improves through retro-action.
Featured Project: AI Macro Agent
This project describes a practical agent loop for companies building AI agents around models, tools, browser/web access, APIs, and feedback.
Agent Capabilities
- Web-aware research before decision-making.
- Tool calling through controlled APIs and backend routes.
- Patch generation for PHP, Python, JavaScript, SQL, and server config.
- Verification using logs, curl tests, syntax checks, and rollback backups.
- Retro-action loop that corrects failed actions instead of blindly continuing.
System Stack
- Frontend: PHP, HTML, CSS, JavaScript.
- Backend: FastAPI, Uvicorn, Nginx, API routing.
- Data: MariaDB integration for real platform data.
- AI: model prompt layer, tool routing, verification, feedback.
- Robotics: AI command interface, DOF planning, simulation pipeline.
Integrated Project Network
My website connects multiple real systems into one portfolio: an AI Agent page, my CV, the Homeless assistance platform, the University platform, and a newspaper homepage redesign demo.
University Platform
Web education platform with users, courses, course material, quizzes, exams, submissions, authentication, and MariaDB-backed data.
Open UniversityHomeless Platform
Assistance-oriented web platform with strategic board, live position data, map visualization, resource tracking, and agent-style support workflows.
Open HomelessNewspaper Demo UI
Responsive editorial homepage redesign for Reflet de Société, embedded locally as a portfolio demonstration with clearer hierarchy and modern navigation.
Open Newspaper DemoWhy This Matters
AI agents need more than text generation. They need access to context, controlled action, memory, validation, recovery, and a clear interface for human review. My work shows this direction through real web systems, deployment patches, database integration, and robotics/agent experiments.