# Firas Ridene > Firas Ridene: AI Engineer based in Tunis, Tunisia, specialized in Generative AI & Multi-Agent Systems. Builds multi-agent systems, real-time voice agents, and AI automation pipelines that ship to production. I'm an AI engineer in Tunis. I build generative AI and multi-agent systems, and I also handle the work around them: talking with the people who'll use the system, finding where their process actually loses time, and turning that into something that ships. That's what I've been doing so far. I built a multi-agent prospecting system that cut manual research by ~80%. VocaFlow is a voice agent platform that handles inquiries, bookings and support across phone, chat and WhatsApp without extra headcount. As a freelancer, I build AI and full-stack solutions for clients in marketing, sales, finance and insurance. I explain every trade-off in plain terms to whoever has to approve it, technical or not. I build agent systems that do real work. Voice agents that answer phones. An agent team that ships code. Pipelines that cut manual prospecting by 80%. I design multi-agent systems end-to-end and pick the stack per problem, not per habit. ## Key facts - Name: Firas Ridene - Role: AI Engineer (Generative AI & Multi-Agent Systems) - Location: Tunis, Tunisia - Availability: Open to opportunities - Languages: Arabic (native), English (fluent), French (fluent) - Education: Engineering Degree in Computer Science, Esprit: School of Engineering (2023 – Oct 2026); Mathematics & Physics, Integrated Preparatory Cycle, Esprit (2020 – 2023) - Shipped projects: 16 · Certifications: 22 - Last updated: 2026-10-20 ## Contact - Website: https://firasridene.tech/ - Email: firas.ridene@outlook.com - GitHub: https://github.com/RideneFiras - LinkedIn: https://www.linkedin.com/in/firas-ridene - YouTube channel: https://www.youtube.com/@FirasRidene - YouTube (project demos playlist): https://youtube.com/playlist?list=PLRjr8ZSeOB_w&si=cJURnWFpjs_r5p9U - Kaggle: https://www.kaggle.com/firasridene - Resume (PDF): https://firasridene.tech/assets/Firas-Ridene-Resume.pdf ## Experience ### AI & Software Engineer · Freelance / Self-Employed (May 2026 – Present) - Delivered AI automation and fullstack solutions for clients across marketing, sales, finance, and insurance sectors. - Translated business needs into technical solutions, working directly with non-technical stakeholders and explaining AI systems in plain terms. - Owned projects end-to-end in full autonomy, from specification through development and deployment. - Stack: Python, React, FastAPI, Supabase, Docker, Azure, CrewAI, Latenode, n8n, Vercel ### AI Engineer (end-of-studies internship) · Agile4UX (Feb – Jul 2026) - Designed a multi-tenant platform where users build and deploy AI voice agents across phone, chat, and WhatsApp. - Built real-time voice agents with knowledge retrieval, data collection, and human handoff at sub-second latency. - Enabled businesses to offer 24/7 customer engagement without added headcount, handling inquiries, bookings, and support. - Stack: React, FastAPI, Supabase, OpenAI Realtime API, Twilio, Stripe, Redis, Docker, Azure ### AI Automation Engineer (internship) · Agile4ux (2025) - Analyzed marketing and sales processes to find repetitive tasks worth automating. - Designed a multi-agent system to research, qualify, and analyze prospects, generating strategic reports. - Cut manual prospecting time by ~80%; automated report delivery and CRM integration. - Stack: Python, FastAPI, OpenAI, CrewAI, Latenode, Presenton ### Data Platform Engineer (internship) · Designet (2024) - Built a platform generating leads from Google Maps with filterable criteria and CSV export. - Reduced research time and improved prospect quality through automated extraction and filtering. - Stack: Google Maps API, SerperAPI, FastAPI, Next.js, Docker ## Case studies ### [Store Support Agent: A support agent that checks real stock and won't be talked into a discount](https://firasridene.tech/work/store-support-agent/) Loomhaus is an AI customer-support agent for a small online store. An orchestrator agent in n8n dispatches to three subagents (inventory, order capture, knowledge base) wired to Google Sheets, Notion and Supabase. The stock decision is made by workflow logic, not by the model, and four independent layers defend against prompt injection. Built in about five hours for the Multi-App AI Agent Hackathon, and tested against the live systems. - Status: Hackathon build · live public demo - Try the live demo: https://loomhaus-agent.vercel.app - Watch the 2-minute demo: https://youtu.be/t3_HilAUTEQ - Source code on GitHub: https://github.com/RideneFiras/MultiAppHackathon - Raw prompt-injection results: https://github.com/RideneFiras/MultiAppHackathon/blob/main/tests/results/02_adversarial_eval.md ### [ChronoWorks: Time, tasks and invoices for freelancers, where the database is the referee](https://firasridene.tech/work/chronoworks/) ChronoWorks is one place where a freelancer runs the business side of their work: clients, projects, time, tasks, leave and invoices, all linked. It is built for Tunisia and France first (TND and EUR, matricule fiscal and SIRET, stamp duty, withholding tax, EU reverse charge). Row-level security, constraints and triggers in Postgres enforce the rules, so a bug in the UI can't produce a wrong invoice or leak another user's data. - Status: Open-source MVP, self-hosted - Source code on GitHub: https://github.com/RideneFiras/ChronoWorks ## Projects ### VocaFlow Multi-tenant platform for building and deploying AI voice agents across phone, chat, and WhatsApp. (voice AI · SaaS) - Designed a multi-tenant platform where users build and deploy AI voice agents across phone, chat, and WhatsApp. - Built real-time voice agents with knowledge retrieval, data collection, and human handoff at sub-second latency. - Enabled businesses to offer 24/7 customer engagement without added headcount, handling inquiries, bookings, and support. - Stack: React, FastAPI, Supabase, OpenAI Realtime API, Twilio, Stripe, Redis, Docker, Azure - Details: https://firasridene.tech/#vocaflow ### 8-Agent Dev Team A full software team staffed by Claude Code agents, running complete sprint workflows solo. (AI-assisted development) - Built a team of 8 specialized Claude Code agents (PM, Frontend, Backend, AI, DevOps, QA, Test, Docs) coordinated through Notion and GitHub. - Ran full sprint workflows solo, from backlog planning to PR review and merge. - Shipped 170+ GitHub issues and 726 automated tests across a ~40,000 line codebase in about 4 weeks. - Stack: Claude Code, Notion MCP, GitHub CLI, Playwright MCP - Details: https://firasridene.tech/#claude-code-team - Source: https://github.com/RideneFiras/ClaudeCodeTeam ### Multi-Agent Prospecting System Agents that research, qualify, and analyze prospects, then write the outreach and the reports. (multi-agent automation) - Designed a multi-agent system to research, qualify, and analyze prospects, generating strategic reports. - Built specialized agents for opportunity analysis, competitor research, and personalized outreach generation. - Reduced manual prospecting time by ~80% and automated report delivery and CRM integration. - Stack: CrewAI, LangChain, FastAPI, Latenode - Details: https://firasridene.tech/#prospecting-system ### Content Creation Automation Automated research, generation, and publication of Facebook content, with a human checkpoint before anything ships. (agent orchestration) - Built a platform automating research, generation, and publication of Facebook content across objectives. - Orchestrated specialized agents per source (news, websites, documents) with email validation before publishing. - Implemented automated comment moderation and intelligent responses based on a knowledge base. - Stack: Python, n8n, MCP, Meta Developer, Tavily, ChromaDB - Details: https://firasridene.tech/#content-automation ### Recruitment Automation System Four interconnected pipelines covering candidate intake, AI scoring, ATS sync, and lead generation. (automation) - Built four interconnected pipelines covering candidate intake, AI scoring, ATS synchronization, and lead generation. - Scored candidates against job requirements using weighted signals, generating clear reasoning behind every shortlist decision and triggering automated shortlist emails and Telegram alerts. - Synced candidate data automatically across the ATS, kanban board, and public job board, keeping every system up to date in real time. - Stack: n8n, Make, REST APIs, Webhooks, Telegram, Notion - Details: https://firasridene.tech/#recruitment-automation ### Airline Customer Service Assistant GenAI capstone (Kaggle × Google): structured-dialogue assistant with RAG over a dedicated knowledge base. (conversational AI) - Designed a conversational assistant that collects customer information through structured dialogue. - Reduced repetitive customer support tasks by automating the handling of frequent inquiries. - Integrated an intelligent search engine to provide contextualized responses from a dedicated knowledge base. - Stack: Python, LangChain, ChromaDB, RAG, Gradio, Gemini - Details: https://firasridene.tech/#airline-assistant - Source: https://github.com/RideneFiras/KagglexGoogle ### Pedagog'IA Social platform for Tunisian teachers with AI generation of fiches, quizzes, and exercises. (EdTech · GenAI) - Built a social platform for Tunisian teachers to share resources, combined with an AI feature generating fiches, quizzes, and exercises. - Implemented an AI credits system, admin panel, and full social feed feature set. - Designed the AI generation flow around subject, level, and chapter inputs for fast lesson prep. - Stack: Next.js, FastAPI, Supabase, pgvector, GitHub Models (gpt-4o-mini) - Details: https://firasridene.tech/#pedagogia - Source: https://github.com/RideneFiras/PedagogIA ### TT Assistant Multilingual voice + text chatbot for Tunisie Telecom. 3rd place at the AI for CSR Hackathon. (conversational AI) - Built an inclusive multilingual chatbot for Tunisie Telecom supporting Arabic, French, English, and Tunisian dialect. - Enabled both voice and text interaction using Whisper for speech and GPT-4o for responses. - Won 3rd place at the AI for CSR Hackathon. - Stack: GPT-4o, Whisper, LangChain, FastAPI, n8n, Telegram - Details: https://firasridene.tech/#tt-assistant - Source: https://github.com/RideneFiras/ChatbotTT ### 5G Throughput Prediction Full MLOps pipeline comparing 300+ models on Azure ML, explainable and monitored in production. (MLOps) - Built a full MLOps pipeline predicting 5G network throughput, comparing 300+ models via Azure ML. - Added explainability with SHAP and GPT-4o to make model decisions interpretable for non-technical stakeholders. - Deployed the pipeline in Docker with monitoring through Elasticsearch and Kibana. - Stack: Azure ML, FastAPI, Docker, Elasticsearch, Kibana, SHAP, MLflow, GitHub Actions - Details: https://firasridene.tech/#5g-throughput - Source: https://github.com/RideneFiras/QosMLOPS ### Industrial Anomaly Detection CNN autoencoder trained only on defect-free screws flags visual defects by reconstruction error. (computer vision) - Built a CNN autoencoder trained only on defect-free images to flag visual defects through reconstruction error. - Delivered real-time inspection with confidence scoring for production line use. - Reduced manual inspection time by 70%. - Stack: TensorFlow, FastAPI, OpenCV, TailwindCSS - Details: https://firasridene.tech/#anomaly-detection - Source: https://github.com/RideneFiras/Anomaly_detection ### Odoo Integrations Odoo CRM wired into a multi-agent prospecting system, plus a facial-recognition attendance module. (integrations) - Integrated Odoo CRM into a multi-agent AI prospecting system for automated lead and pipeline updates. - Built a facial recognition attendance system integrated with Odoo's attendance module. - Stack: Odoo, Python, OpenCV - Details: https://firasridene.tech/#odoo-integrations ### LeadGenMAP Platform that generates qualified leads from Google Maps with filterable criteria and CSV export. (lead generation) - Built a platform that generates leads from Google Maps with filterable search criteria and CSV export. - Reduced research time and improved prospect quality through automated extraction and filtering. - Stack: Google Maps API, SerperAPI, FastAPI, Next.js, Docker - Details: https://firasridene.tech/#leadgenmap - Source: https://github.com/RideneFiras/LeadGenMAP ### n8n Workflow Library Public collection of the n8n workflows behind my automation work, packaged for reuse. (automation) - Open-sourced production workflows including Facebook sales conversation automation and a hiring agent. - Includes facial recognition and tender processing flows built as reusable n8n templates. - Stack: n8n, Webhooks, REST APIs - Details: https://firasridene.tech/#n8n-workflows - Source: https://github.com/RideneFiras/Myn8nWorkflows ### Churn Prediction & Billing Analysis, Telecom Subscriber Data A predictive analytics project uncovering key churn drivers and revenue patterns across 20K telecom subscribers to power targeted retention strategies. (Data science) - Analyzed 20k subscriber records to uncover churn drivers, retention dynamics, and revenue patterns. - Built cohort and monthly churn analysis, identifying early-tenure churn segmented by technology and acquisition channel. - Linked billing patterns to retention behavior and proposed a churn prediction framework to prioritize retention actions. - Stack: Python, Pandas, NumPy, Matplotlib - Details: https://firasridene.tech/#churn-prediction-billing-analysis-telecom-subscriber-data ### Store Support Agent AI customer support agent for e-commerce that checks real inventory before promising availability, and resists prompt-injection attempts to override its own rules. (AI Automation) - Built a multi-agent support system where an orchestrator dispatches to RAG, inventory, and order-capture subagents, syncing live across Notion and Google Sheets. - Handled the overselling edge case deterministically, refusing to confirm stock that's already pending on another order. - Defended against prompt injection with no discount/price tools in the system, a pre-filter guard, and a structured-output schema as a final backstop, tested against a set of adversarial prompts. - Stack: n8n, Notion, Google Sheets, Supabase, pgvector, Claude, Next.js, Vercel - Case study: https://firasridene.tech/work/store-support-agent/ - Source: https://github.com/RideneFiras/MultiAppHackathon - Demo: https://youtu.be/t3_HilAUTEQ ### ChronoWorks Unified platform for freelancers to manage clients, projects, time, tasks, leave, and invoices in one linked system. (SaaS · freelance ops) - Designed a linked data model where logging time on a task feeds project status, leave-adjusted totals, and client invoices automatically. - Built a locked invoicing system with gapless numbering and immutable issued invoices, enforced with Postgres triggers and row-level security. - Implemented tax-profile logic for Tunisia and France, covering VAT, matricule fiscal, stamp duty, and reverse charge. - Stack: Next.js, TypeScript, Supabase, PostgreSQL, Tailwind CSS, dnd-kit, Vercel - Case study: https://firasridene.tech/work/chronoworks/ - Source: https://github.com/RideneFiras/ChronoWorks ## Skills - Generative AI & Agents: LLMs, RAG, LangChain, CrewAI, MCP, Claude Code, Transformers - ML & MLOps: PyTorch, TensorFlow, Scikit-learn, MLflow, Azure ML Studio, SHAP, GitHub Actions - Development & Data: Python, JS (Node.js), Java (Spring Boot), FastAPI, REST APIs, Pandas, SQL, NoSQL - Infra & Automation: Docker, Render, Git/GitHub, Elasticsearch, Kibana, n8n, Latenode ## Certifications - AI Security & GenAI Red Teaming (ESPRIT) - Attendance Hashgraph Developer (Hedera) - AWS Academy Graduate: Cloud Foundations (Amazon Web Services) - Applications of AI for Anomaly Detection (NVIDIA) - Building Generative AI-Powered Applications with Python (IBM) - Multi AI Agent Systems with crewAI (CrewAI) - Microsoft Azure AI Essentials Professional Certificate (Microsoft) - Building AI Agents with Multimodal Models (NVIDIA) - Generative AI with Diffusion Models (NVIDIA) - Machine Learning in Production (DeepLearning.AI) - Building LLM Applications With Prompt Engineering (NVIDIA) - Introduction to Transformer-Based Natural Language Processing (NVIDIA) - 5-Day Gen AI Intensive (Kaggle) - Building RAG Agents with LLMs (NVIDIA) - Machine Learning Explainability (Kaggle) - Fundamentals of Deep Learning (NVIDIA) - GitHub Foundations (GitHub) - CCNA: Switching, Routing, and Wireless Essentials (Cisco) - Hands-on with Explainable AI Tools & Techniques (ESPRIT) - Artificial Intelligence Fundamentals with Capstone Project (IBM) - Agentic AI for Enterprise (ESPRIT) - Data Engineering & MLOps (ESPRIT) ## Optional - [Book a call (15-min introduction or 30-min AI & automation discussion)](https://firasridene.tech/book/) - [Legal notice, terms of service & privacy policy (French)](https://firasridene.tech/legal/) - [Website source code](https://github.com/RideneFiras/RideneFiras.github.io)