[nom]

AI & Backend Engineer | Production LLM & MLOps

Based in Morocco | Open to Remote & International Opportunities

[e-mail] | [lien] | [téléphone]

PROFESSIONAL SUMMARY

AI Engineer & Backend Engineer with 4+ years of experience building and scaling production-ready LLM and ML systems. Specialized in model training, fine-tuning, RAG pipelines, and multi-agent orchestration, with a strong backend foundation in FastAPI, vector databases, and MLOps. Experienced in delivering secure, high-performance AI services from architecture to production.

PROFESSIONAL EXPERIENCE

Freelance AI Engineer – Satellite Computer Vision

GeoAP · Dec 2025 – Present | Remote, Morocco

• End-to-End Satellite CV Pipeline: Architected and delivered a production-grade MLOps pipeline for detection and semantic segmentation of agricultural and urban objects from satellite imagery, processing 180 source images into 12,000+ training tiles via overlap-aware tiling with zero data leakage across train/val/test splits.
• Multi-Model Architecture: Trained three complementary models — YOLO11m object detection (mAP50 0.836), YOLO11m instance segmentation (mAP50-mask 0.737), and fine-tuned SAM1 ViT-H semantic segmentation (mIoU 0.804) — covering 10 object classes including buildings, trees, solar panels, swimming pools, cultivated parcels, rivers, and roads.
• SAM1 Fine-Tuning: Fine-tuned SAM1 ViT-H mask decoder (4M trainable params, 300M-param encoder frozen) with pre-computed embedding caching across 9,958 tiles, reducing epoch time from 88 min to 5 min on a T4 GPU.

Implemented custom Dice + Focal loss and torch.amp mixed-precision training.

• GIS-Ready Output: Inference results exported as GeoJSON FeatureCollections with real-world WGS84 coor- dinates, enabling detected objects and segmentation masks to be visualised directly on real maps in Google Earth, QGIS, Leaflet, and any GIS-compatible tool.
• SAHI Tiled Inference: Integrated Slicing Aided Hyper Inference (SAHI) for detection on high-resolution satellite imagery, enabling robust small-object detection across arbitrary image resolutions.
• FastAPI Inference Service: Built a FastAPI REST API with four inference endpoints, lazy singleton model loading, and an interactive dark-theme frontend with live legend and GeoJSON download.
• MLOps & CI/CD: Configured a full DVC pipeline DAG (12 reproducible stages) with Roboflow dataset versioning.

Implemented GitHub Actions CI/CD with linting, 25 unit tests, API smoke tests, and Docker build validation on every push.

• Tech Stack: PyTorch · YOLO11m · SAM1 ViT-H · SAHI · FastAPI · rasterio · Shapely · OpenCV · DVC · Roboflow · Lightning AI (T4 GPU) · Docker · GitHub Actions AI Engineer / Software Engineer Assure Solutions – Client: MAMDA-MCMA (Major Insurance Group) Sep 2023 – Aug 2026 | Rabat & Kenitra, Morocco
• AI-Powered Document Automation Platform: Architected and deployed an intelligent document processing system integrating OCR, NLP, and vector search for insurance workflows.
• Hybrid OCR Engine: Developed a Tesseract + EasyOCR pipeline achieving 91.3% accuracy on Arabic/French documents with intelligent text extraction and reconstruction.
• Semantic Search Engine: Built Pgvector vector database with HNSW indexing enabling sub-second semantic retrieval across 760+ insurance documents, improving information access by >40%.
• LLM Integration: Implemented Phi-3 LLM via Ollama for local Named Entity Recognition and document classifi- cation with >90% precision, automating manual categorization.
• Secure Infrastructure: Deployed S3 storage with AES-256 encryption for GDPR compliance and full audit trails, implementing Zero-Trust security principles.
• Full-Stack Interface: Delivered web application for document upload, semantic search, and validation, integrating ChatBot Assistant for contextual queries and summarization.
• Insurance System Automation: Engineered automation for RENP system (Recovery of Unpaid Installments) and formal notice generation using JasperReports, streamlining client communication.
• Performance Optimization: Optimized SQL Server stored procedures and resolved critical production bugs, reduc-

1

ing system latency by ~30%.

Software Engineer

Assure Solutions · Aug 2023 – Sep 2026 | Kenitra, Morocco

• Developed and maintained core insurance modules (Client, Victim, Sinister Management) using PHP Phalcon and SQL Server.
• Designed and implemented a secure RBAC/PBAC system for 11 distinct user profiles, controlling access to sensitive data and features.
• Architected a Document Management System (GED) using MinIO for secure storage and retrieval of insurance docu- ments.
• Integrated OTP authentication to strengthen security for user authorization flows.

SELECTED AI & BACKEND PROJECTS

Premium Secure AI Backend (Zero-Trust LLM Infrastructure)

Personal Project / Production Blueprint

• Architected a production-grade, GPU-aware AI inference backend unifying OpenAI and local Ollama models ( Phi , Llama-2 , Mistral ) behind a [information personnelle] hardened FastAPI service.
• Implemented enterprise-grade security: Argon2-id password hashing (GPU-resistant), short-lived scoped JWTs, LLM firewall for pre-inference prompt injection/PII blocking, and Redis based rate limiting (requests, tokens, GPU- seconds).
• Designed full audit trails with structured JSON logging for compliance (SOC 2, GDPR) and built a Docker Compose stack for one-command deployment (API, Redis, Ollama).
• Demonstrated Impact: Secured AI service pattern where stolen credentials/tokens yield minimal, auditable access directly applicable to regulated healthtech/fintech AI agent roles.

M3AE – Multi-Modal Medical Vision & Language Model

Research Project | PyTorch , Transformers , Multi-Modal Language

• Developed a Transformer-based multi-modal architecture using Masked Autoencoders for joint understanding of medical images and clinical text.
• Pre-trained on the ROCO dataset (81k radiology image-text pairs) and fine-tuned on VQA-RAD, achieving state-of-the-art results on medical visual question answering.
• Applied deep learning techniques: CNN feature extraction, Transformer attention mechanisms, and contrastive learning for vision-language alignment.

AI-Powered Chatbot (RAG)

Professional Project | FastAPI , ChromaDB , LangChain , React

• Implemented a Retrieval Augmented Generation RAG chatbot using HuggingFace LLMs (Flan-T5, BART) and ChromaDB.
• Utilized LangChain for orchestration, enabling context-aware conversations and document-based Q&A.
• Built an interactive React frontend for real-time chat visualization and seamless user interaction.

Cryptocurrency Forecasting & Automated Trading System

Personal Project | Python , Scikit-learn , XGBoost , LSTM , Binance API

• Created a ML-powered forecasting pipeline using ARIMA, XGBoost, Random Forest, and LSTM models for ETH/USDT prediction.
• Built a custom backtesting framework with dynamic risk controls, achieving 5.41% ROI and MAE of 0.0065 in volatile market conditions.
• Integrated classical ML and deep learning approaches with real-time market data and sentiment analysis (via LLMs) into an n8n automated workflow.

CORE TECHNICAL EXPERTISE

AI & LLM Engineering

• Production LLM Systems: RAG pipelines, agent orchestration ( LangChain ), evaluation frameworks, prompt engineering, fine-tuning/ LoRA

2

• Model Integration: OpenAI API, Ollama ( Phi , Llama , Mistral ), AWS Bedrock, Google Gemini
• Vector Databases: ChromaDB, Pinecone, pgvector
• Agent Systems: Multi-agent workflows, tool calling, MCP, A2A architectures, autonomous reasoning
• Trust & Safety: LLM firewalls, hallucination control, audit logging, Zero-Trust security Machine Learning & Deep Learning
• Architectures: Transformers, CNNs, RNNs/LSTMs, Masked Autoencoders, BERT-based models
• Frameworks: PyTorch , TensorFlow , Scikit-learn , Pandas , NumPy , Keras
• Classical ML: SVM, Decision Trees, K-Means, PCA, XGBoost, ensemble methods
• Lifecycle: Research to production training, validation, optimization, deployment
• Computer Vision: OpenCV , YOLO11 , SAM , SAHI , OCR ( Tesseract , PaddleOCR , EasyOCR ), VLM , satellite imagery analysis Backend & Cloud Engineering
• Languages: Python (Advanced), TypeScript / JavaScript , PHP , Java
• Frameworks: FastAPI , Flask , NestJS , Spring , Laravel / Phalcon
• APIs: REST, gRPC, WebSockets, OAuth2/JWT, RBAC/PBAC
• Databases: PostgreSQL , SQL Server , MySQL , Redis , MongoDB
• DevOps & MLOps: Docker , Kubernetes , DVC , CI/CD ( GitLab CI , GitHub Actions ), Lightning AI
• Observability: logging, monitoring, performance profiling
• Workflow Automation: n8n , workflow orchestration, ERP/CRM integration

CERTIFICATIONS

• Artificial Intelligence on Microsoft Azure — Microsoft (Coursera), March 2026
• Indexing, Performance Optimization & Functions in SQL Server — Microsoft (Coursera), March 2026
• AI for Business Professionals — HP LIFE (HP Foundation), April 2026

EDUCATION

• Master’s Degree in Artificial Intelligence & Virtual Reality: Ibn Tofail University, Morocco (2025)
• Bachelor’s Degree in Computer Science & Mathematics: Ibn Tofail University, Morocco (2023)

LANGUAGES

• Arabic (Native), English (Professional), French (Professional)

3