AI Engineer in France — 2+ years building production LLM agents, multi-agent systems, and RAG pipelines that ship to regulated domains.

• 🔭 2+ years building AI systems — across academic research (Master in Big Data & AI at Paris-Dauphine, Lauréate), freelance, and full-time roles since 2023.
• 🤖 Shipping agentic AI copilotes at TerraGrow — multi-agent LLM orchestration, RAG, voice (Whisper Large v3), OCR (Azure Document Intelligence), satellite imagery, and MLOps from day one.
• 📚 Going deeper on: inference optimisation (vLLM, quantisation), eval frameworks for multi-agent systems, satellite imagery × LLM pipelines (NDVI / GNDVI).

AI Engineer · TerraGrow · Mar 2026 – Present Scaling Cowgito, a multimodal agentic AI copilot for agricultural operations. Designing multi-agent LLM systems with RAG pipelines for contextual reasoning. Building autonomous agents that orchestrate complex business workflows. Industrialising NDVI / GNDVI satellite-imagery pipelines for crop-health monitoring. Production-grade MLOps / DevOps (CI/CD, cloud, monitoring). Stack: Python · TypeScript · LLMs · LangChain · RAG · Multi-Agent Systems · CI/CD · MLOps

AI Engineer (Freelance) · TerraGrow · Sep 2025 – Present Continuous collaboration on Cowgito. Multi-agent LLM workflows with RAG for document processing and retrieval. NDVI / GNDVI satellite pipelines. DevOps / CI-CD for quality, maintainability, and scalability. Stack: Python · TypeScript · LLMs · LangChain · RAG · Multi-Agent Systems · CI/CD

AI Engineering Intern (AgriTech) · TerraGrow · Feb 2025 – Sep 2025 Designed and shipped the first version of Cowgito — multimodal agentic copilot with text / voice / vision interactions. Built a multi-tenant RAG system for agricultural document management. Industrialised the NDVI / GNDVI image-processing pipeline. Established a DevOps / CI-CD practice for AI solution quality. Stack: Python · TypeScript · LLMs · LangChain · RAG · Multi-Agent Systems · CI/CD · SonarQube · Grafana

Project Mentor · Quebec Training Center · Jun 2023 – Oct 2023 Selected post-internship to mentor licence students on AI and web-development projects. Leadership and project management.

Full-Stack Developer · Quebec Training Center · Feb 2023 – May 2023 Built and maintained websites, web applications, and APIs for multiple clients. Stack: Django · PostgreSQL · React.js · three.js · NLP · Web Security

Languages: Python · TypeScript · SQL · Bash ML / AI: PyTorch · TensorFlow · LangChain · LangGraph · Hugging Face Transformers · Whisper Large v3 · GPT-4 / 4.1-mini · OpenCV · YOLOv8 RAG & Vector: Qdrant · FAISS · pgvector · Azure Document Intelligence (OCR) Backend: FastAPI · SQLAlchemy (async) · PostgreSQL · Redis · Celery · ReportLab · Asynchronous Programming Frontend: React · Vite · TanStack Query · Zustand · TailwindCSS Data & Monitoring: AI Observability · Grafana · Prometheus · ETL · Multivariate Analysis (PCA, MCA, K-means, Ward) Web Scraping: BeautifulSoup · Scrapy · Selenium Infra & MLOps: Docker · Docker Compose · GitLab CI/CD · GitHub Actions · SonarQube · MinIO Cloud: Azure · AWS · GCP · OVHcloud

A regulated-domain portfolio of 17 production-shaped AI products. Each ships a real cross-modal AI workflow (voice → analyzer → cross-validation → PDF brief). Private repos — request access for a walkthrough.

• investigateiq-ai — Voice-to-document corroboration for investigative journalism: extracts entities from a source interview, cross-checks against every case document, surfaces ranked story leads + chase-list.
• pharmavig-ai — Pharmacovigilance signal detection: MedDRA auto-coding, PRR + ROR + χ² disproportionality, ICH-E2B(R3) expedited 7/15-day drift detection.
• auditlens-ai — Financial audit (PCAOB AS 2315): Benford's Law chi-square + offshore money-flow graph + PCAOB AS 2401 management-override detection.
• claimiq-ai — P&C insurance claim fraud: 6-taxonomy filterable indicators with per-indicator confidence + voice ↔ document cross-modal analyser.
• bidiq-ai — AI-drafted FAR-compliant RFI letters from subcontractor bids, with scope-gap detection at unit-rate level.
• receiveiq-ai — GPT-4 VISION multimodal warehouse-damage assessment with Incoterms-aware liability reasoning.

Plus 11 more across customs (HS code cross-validation), HR (procedural compliance grading), govtech (SNAP/Medicaid eligibility), agriculture (N-P-K + IPM precision-ag), real estate (commercial lease clause analysis), water utility (permit allocation tracking), and building IoT (ASHRAE voice + sensor fusion).

• Cowgito — Multimodal Agentic AI Copilot (TerraGrow). End-to-end design and deployment: natural-language complex-query handling, internal-data integration (API, SQL, RAG), specialist-agent orchestration, scalable cloud deployment with performance monitoring.
• Elper — Multi-Tenant RAG Platform for Document Management (TerraGrow). Secure multi-tenant document RAG with Qdrant vector search and MinIO + PostgreSQL persistence.
• NDVI / GNDVI Satellite-Imagery Pipeline (TerraGrow). Industrialised a prototype into a robust crop-health monitoring tool with vegetation-index analysis and per-parcel tracking.
• Churn Prediction Model (Minerva Training). Random Forest at 93 % accuracy, deployed via a Flask application.

• Research on Student Learning Preferences (Université). PCA, MCA, K-means, and Ward classification on questionnaire responses to surface learning-style clusters.
• Web Scraping & E-commerce Recommendation System (Université). Python scraper over Tunisian e-commerce sites, MySQL storage, visualisation, recommendation engine.

• 🥉 3rd place — Airbnb UX Hackathon (Paris 2024 Olympics edition) — geographic analysis + NLP + host-profile study to improve UX for the Games.
• 🥉 3rd most innovative project 2023 — Adaptive Language Test-Prep Platform (Quebec training center). Django + Flask + React + PostgreSQL + chatbot + 3D visualisation + security audit.
• Hospital HR AI System (Hackathon). ML models to monitor patient state and predict nosocomial-infection risk in real time.
• Top 5 % — Kaggle House Prices — advanced regression + blending (rank 228 / 4841).
• Alzheimer's detection by handwriting analysis (Kaggle). ResNet CNN at 0.88 precision.

• RAG + Google Gemini chatbot on GCP Cloud Run (Université). LangChain + FastAPI + Google Cloud SQL, Dockerised and secured.
• BERT-fine-tuned chatbot for cyberattack news (Université). NLP + fine-tuned BERT for real-time cybersecurity-news delivery.

• Master en Big Data & Intelligence Artificielle — Université Paris-Dauphine (PSL) · 2023–2025 · Mention bien, Lauréate Core: Maths & Optimisation · Programming & Algorithms · Data Analysis & Modelling · ML / Deep Learning / RL / NLP / Generative AI · Big Data Systems · Graph Theory
• Licence en Sciences de l'Informatique — Institut Supérieur des Arts Multimédia de la Manouba · 2022–2023 · Mention très bien, Lauréate Core: Web & Mobile Dev · Programming Languages · Image Processing & AI · VR/AR · Databases · Networks & Security · Cloud & Big Data
• Baccalauréat en Science Expérimentale — Lycée Ibn Sina, Tunis · 2019–2020 · Mention assez bien

• 🔗 LinkedIn: linkedin.com/in/wissal-ben-othmen-2231521bb
• 📧 Email: wissal.benothmen01@gmail.com
• 🤝 GitHub: @wissalbenothmen

_{🇫🇷 Based in France · Languages: French — Intermediate Advanced · English — Professional}