Safinez Hlila

Full project ownership (architecture → development → testing → deployment): Python, FastAPI, React, Node.js • Bulk CV parsing with Mistral/Ollama: 10,000+ resumes processed in minutes with structured extraction of skills, experience and education • AI matching engine: SBERT embeddings + Elasticsearch 8.x for precise candidate-to-job matching based on business criteria • Automated communication agent: email qualification, follow-ups and profile updates without human intervention — Zoho CRM & Google Drive integration • Interview orchestration agent: automated link delivery → timed session with fraud detection → result collection → automatic profile update • API testing (unit, integration) + security testing (injection, XSS, DoS) — Dockerized deployment on Ugreen NAS 4800DXP, running stable in production for 4 months ▶ Automated Monitoring, Sourcing & Marketing Agents • Autonomous APEC monitoring agent: weekly scraping (Playwright + Python), normalization and synchronization of job offers with active hiring needs • Daily HiTechPro sourcing agent: automatic candidate database enrichment + scoring and matching against open positions • LinkedIn marketing agent: content generation → visual creation → scheduled publishing according to editorial calendar — all containerized Docker/NAS ▶ IT Market Intelligence Platform — Flask + Mistral AI + n8n • Flask REST API scraping Turnover IT & HiTechPro (Selenium headless) — Mistral AI for semantic job classification, ReportLab for automated PDF reports, n8n for scheduled pipeline orchestration

Designed and developed a CNN object recognition model (TensorFlow/Keras + OpenCV) — built and labeled the training dataset, applied Transfer Learning to accelerate convergence • Image preprocessing pipeline: resizing, normalization, augmentation (rotation, flipping, brightness) — hyperparameter tuning, rigorous evaluation (mAP, precision, recall, F1) and results presented to the team • Agile Scrum team member — daily standups, sprint planning, retrospectives

Conception et réalisation d'une analyse des sentiments sur les réseaux sociaux utilisant des algorithmes de traitement du langage naturel (NLP) et des modèles d'apprentissage automatique, intégrant Power BI pour l'interprétation des résultats.

Received raw data via internal Veeva CRM exports and CSV files — multi-step processing: collection → cleaning → deduplication → validation → structured delivery back into the CRM • Anomaly detection across large datasets (missing fields, duplicates, inconsistent formats) — batches validated by manager before CRM import · data governance documentation maintained

Développement d'une application web _ mobile

Développement d'une application web interne à l'entreprise avec la technologie mern Stack

Développement d'une application web Cowork-Space

Créations des maquettes et des affiches.