Hey, I'm Lojaine

I build ML systems that actually work.

AI EngineerComputer VisionMLOps

Building production AI systems, from detection pipelines to deployed dashboards.

About

AI Engineer with an Engineering Degree in AI & Big Data (EUR-ACE accredited, ISGA Marrakech) and an MSc in Data Engineering & Cloud Computing (Aivancity School for Technology, Business & Society, Paris-Cachan).

I specialize in computer vision and applied deep learning, with end-to-end production experience across detection pipelines, forecasting models, and full-stack AI systems—from YOLO-based object detection to FastAPI backends and the dashboards and apps built on top of them.

Most recently, I built and deployed a live intelligent queue management system running in a real retail environment, combining computer vision, ensemble forecasting, and a bilingual mobile app used by store staff.

Experience

AI / Computer Vision Engineer — Final Year Internship

Feb 2026 - Jun 2026

Roc4Tech / VSENS • Remote, UAE — System deployed in a retail store, France

  • Trained a Prophet + LSTM + XGBoost ensemble on 183,965 real customer events, hitting sub-6-minute MAE on short-horizon queue predictions across a live 4-lane checkout
  • Built real-time CV pipeline on two live camera feeds (entrance + checkout/exit): a YOLOv9 head-detection model (OpenVINO-optimized) for people counting, paired with OWLv2 for trolley/basket equipment classification, monitoring up to 4 checkout lane slots via RTSP
  • Designed a FastAPI + PostgreSQL/TimescaleDB backend powering a Streamlit ops dashboard and a React Native mobile app
  • Shipped a React Native/Expo supervisor app with live queue status, 60-min/3-hour forecasts, daily recap, and bilingual EN/FR support
  • Deployed the full system to production with automation and Cloudflare Tunnel for remote access, running continuous auto-prediction

AI / Software Engineer — Part-time

Nov 2025 - Jun 2026

Machine-Vision.io (Inetech.io) • Remote, Germany

  • Fine-tuned YOLOv8n for tyre inspection: surface defect detection, damaged tyre classification, and production-line presence detection
  • Integrated PLC-triggered industrial cameras at production-line barriers, capturing tyre images on the fly to feed the detection pipeline
  • Led a solar panel energy optimisation project end to end, from data processing through model deployment
  • Built and maintained internal web tools with Nuxt.js/Vue.js
  • Contributed to CV model development, data pipeline design, and software engineering across multiple product workstreams

Computer Vision Engineer — Internship

Jul 2025 - Aug 2025

Machine-Vision.io (Inetech.io) • Remote, Germany

  • Built production OCR pipelines for European vehicle registration documents, processing 1,000+ docs/day with a 35% accuracy improvement
  • Developed and integrated computer vision modules for document understanding and automated data extraction

Data Analyst — Internship

Jul 2024 - Aug 2024

Terrain 360 • Marrakech, Morocco — On-site

  • Designed and orchestrated data pipelines for cleaning, transforming, and modelling raw geospatial imagery data
  • Integrated processed data into an interactive analytics dashboard to support strategic decision-making

Insurance Operations Intern

Jul 2023 - Aug 2023

Allianz Maroc • Marrakech, Morocco — On-site

  • KPI dashboards, data reporting, and operational analysis across insurance product lines

Education

MSc Data Engineering & Cloud Computing

2024 - 2026

Aivancity School for Technology, Business & Society • Paris-Cachan, France

Engineering Degree — AI & Big Data (EUR-ACE accredited)

2021 - 2026

ISGA Marrakech • Marrakech, Morocco — Dual degree programme

Projects

Industry Projects

Intelligent Queue Management System

Feb 2026 – Jun 2026 • Roc4Tech / VSENS Final Year Internship — Deployed at a retail store, France

Live, production computer vision and forecasting system for retail queue management, running on two active camera feeds — an entrance camera and a checkout/exit camera monitoring up to 4 lane slots. Real-time detection runs a YOLOv9 head-detection model (OpenVINO-optimized) for people counting, with OWLv2 classifying trolley/basket equipment. Paired with a Prophet + LSTM + XGBoost forecasting ensemble trained on 183,965 real customer events, achieving sub-6-minute MAE on short-horizon queue predictions. Delivered as a FastAPI + PostgreSQL/TimescaleDB backend powering a Streamlit ops dashboard (live lane monitoring, forecast charts, alerts, demographic analytics) and a bilingual EN/FR React Native/Expo supervisor app, deployed with Cloudflare Tunnel for remote access and continuous auto-prediction.

YOLOv9OWLv2OpenVINOProphetLSTMXGBoostFastAPITimescaleDBReact Native
Confidential industry project

Tire Pattern Classification System

Summer 2024 • Inetech.io Internship

Production-ready deep learning system for tire pattern recognition. Trained and compared 5 architectures (EfficientNet, ResNet, ConvNeXt, ViT, custom CNN) with engineered augmentation pipeline. Processed 33,639 records across 173 brands and 2,328 unique patterns.

PyTorchFastAPIDockerOpenCVAlbumentations
Industry Project

OCR Vehicle Registration System

Summer 2025 • Inetech.io Internship

Production OCR system processing 1000+ German vehicle registration documents daily with 99.2% uptime. Implemented computer vision pipeline with OpenCV and custom filtering algorithms, deployed with Docker and GCP.

OpenCVFastAPIDockerGCPGoogle Gemini

Personal Projects

Weather Forecasting with LSTM

LSTM-based 7-day weather prediction achieving 87% accuracy. Processed 2 years of multi-source meteorological data with temporal feature engineering.

TensorFlowLSTMPython
View on GitHub

Bank Fraud Detection

Led 3-person team building ML fraud detection for Moroccan banking sector. Analyzed 284,807 transactions with heavy class imbalance, reducing false positives by 23%.

Scikit-learnPandasXGBoost

Smart IoT Home System

Home automation system with 15+ sensors deployed in real environment, achieving 30% energy consumption reduction. Full-stack solution from hardware to web dashboard.

ArduinoESP32ReactNode.js

Smart City IoT Sensor Data Warehouse

Data warehouse and analytics platform for city sensor networks. Star-schema design (fact readings joined against sensor, location, time, and weather dimensions) with an ETL pipeline that cleans raw readings, flags anomalies, and powers analytics on traffic, noise, and air quality by district.

PostgreSQLPythonPandasDocker
View on GitHub

Social Network Influence Analytics

Real-time big data pipeline analyzing influence and recommendation networks on social media. Kafka streams user interactions into a Bronze/Silver/Gold data lake via Spark, feeding a Neo4j graph for PageRank-based influencer ranking and Louvain community detection, visualized in Streamlit.

KafkaSparkNeo4jStreamlit
View on GitHub

E-Commerce Data Warehouse & BI

End-to-end business intelligence system for online sales analysis. Star-schema warehouse in PostgreSQL, automated ETL pipelines built in Pentaho Data Integration, and interactive Power BI dashboards covering sales trends, top products, and customer geography.

PostgreSQLPentaho ETLPower BI
View on GitHub

Skills

Computer Vision

YOLOv8/v9/v11OWLv2OpenVINOOpenCVByteTrackRTSP pipelinesOCRFFmpegPLC-triggered camera integration

ML / AI

TensorFlowPyTorchProphetXGBoostScikit-learnDVCLSTMEnsemble methods

Big Data & Backend

FastAPIPostgreSQLTimescaleDBKafkaSparkPySparkStructured StreamingDelta LakeNeo4jDockerREST API

Frontend

ReactReact NativeExpoNuxt.jsVue.jsStreamlit

Cloud / Infra

AWS (ECS, EC2)GCPAzureAzure Digital TwinsCloudflare Tunnel

BI / Analytics

Power BIPentaho Data IntegrationStar-schema data warehousing

Languages

PythonJavaScript/TypeScriptSQLBash

Let's Connect

Open to internships, ML collaborations, and production ML projects

Marrakech, Morocco

Arabic (Native) • French (Fluent) • English (Fluent) • Russian (Intermediate)