Hi, I'm
I build full-stack web & mobile apps with a Python backend and an AI core, language models, NLP and computer vision. Comfortable taking a project from the first commit all the way to the server it runs on.
I hold a Master's in Data Science from the Faculty of Sciences of Monastir, Tunisia, and a Bachelor's in Software Engineering & Information Systems. Over the past two years I've shipped several applications that pair a web or mobile frontend with a Python backend and a real AI component.
I like owning the whole stack: designing the API, training or wiring up the model, building the UI, and hardening the Linux server it all runs on (Docker, nginx, UFW, fail2ban). Most of my apps are localized in English, French and Arabic (with full RTL support).
Highlighted = my core stack, the tools I use every week.
A project board where every task is worked on by a real Claude Code agent: it writes a plan, you approve it, it works in an isolated git worktree, the checks run and a reviewer agent reads the diff. Nothing reaches the code until you accept it. FastAPI backend, installable web app.
A YouTube channel that teaches English to children aged 3–6, produced by a pipeline: lip-synced 2D animation rendered with Remotion, text-to-speech voices checked by speech recognition, and n8n orchestration with Telegram approvals before every upload through the YouTube Data API.
Mobile app that takes a photo of you and a garment and generates a photorealistic try-on image with hosted diffusion models (Leffa by default, IDM-VTON as fallback), then turns the result into a rotatable 3D model. React Native (Expo) client, FastAPI backend.
Multi-tenant SaaS where each client builds an AI assistant trained on its own documents and crawled websites, using a RAG pipeline (sentence-transformers + ChromaDB). Built with a FastAPI & MongoDB backend, a React admin dashboard, an embeddable chat widget and a Celery crawler.
Full mobile fitness app: workout sessions with per-set weight/rep tracking, rest timer and automatic personal-record detection (seeded with ~83 exercises). Step counting via HealthKit & CoreMotion on iOS and the hardware sensor on Android. Hardened, self-managed VPS deploy.
A chatbot that turns plain-English questions into T-SQL, runs them on an MS SQL Server database and answers in a sentence. Flask back end with pyodbc; one LLM writes the SQL (DeepSeek via OpenRouter), another phrases the answer (Mixtral on Groq).
2025Web app to centrally manage multiple Discord bots and servers from a real-time admin dashboard. AI agents generate natural, context-aware replies and post them through Discord webhooks.
The monitoring I run on my own servers: Bash scripts on cron that send Telegram alerts for never-seen SSH logins, revoked keys, changes to authorized_keys, services down and tampered GitHub repositories — plus a portable Python rewrite (standard library only, systemd-hardened) with /status commands.
2024LLaMA-2-powered chatbot that helps users find IT products with personalized recommendations, kept fresh by a daily web scraper pulling the MyTek catalog.
2024Web app integrating BART for summarization and Helsinki-NLP for translation, faster summaries with support for 10+ languages.
2024Trained a YOLO model to detect and extract vehicle license plates from video, frame by frame, in real time.
I work as a freelance full-stack & AI developer based in Tunisia (UTC+1) and I'm also open to full-time roles. Remote-first, in English, French or Arabic — from a first prototype to an app running in production.
An assistant that answers from your own documents and website instead of guessing.
React and React Native front ends on a FastAPI or Node.js back end, designed, built and shipped.
Telegram and Discord bots, scrapers and scheduled jobs that run unattended and tell you when something breaks.
I put the app in production on a Linux server and keep it fast, patched and locked down.
Master's final-year project (PFE)
Final-year project (PFE): a multi-tenant AI chatbot platform where admins create assistants trained on their own content, uploaded documents and crawled websites, via a RAG pipeline (sentence-transformers embeddings + ChromaDB). Built the FastAPI + MongoDB backend, a React admin dashboard, an embeddable chat widget, and a Celery/Redis/Playwright crawler, with JWT + 2FA auth, rate limiting, and Docker deployment.
Case studyMonastir, Tunisia
Six-month internship across two projects: a React Native mobile app + web dashboard for secure document exchange between accountants and clients, and a business-address platform (graduation project) letting companies register a legal office without renting a physical space, React.js frontend, Node.js backend.
Faculty of Sciences of Monastir, Tunisia
Faculty of Sciences of Monastir, Tunisia
Dhouiher High School, Zarzis, Tunisia
Write-ups of real incidents and fixes from my own servers and projects — what broke, why, and the exact commands that fixed it.
A hidden folderOpen task, a "font" that is really JavaScript, and commits rewritten with their original timestamps. What the infection looks like from the inside, the grep that finds it, and the cleanup mistakes I made so you don't have to.
Read articleUFW was active, default deny, and a Postgres and a Redis container had still been reachable from the internet for six weeks. Why Docker's iptables rules win, how to check in 10 seconds, and the three ways to fix it.
Read articleThe checkbox stays greyed out, the certificate never arrives, and http:// keeps serving your site without a redirect. The DNS mistake that causes the loop, how to read the real certificate state from the API, and the remove-and-re-add trick that works.
Read articleI'm open to full-time roles and freelance projects. Write to me here or by email — I usually reply within a day.
or reach me directly
ouiriemmiraed@gmail.com
