Saad Yaqine
ML Engineer & Data Scientist
I build production-ready AI systems that solve real business problems. Specializing in NLP, LLMs, and end-to-end MLOps pipelines, I've deployed sentiment analysis models for financial trading, automated code review systems, and intelligent document retrieval platforms. From data streaming architecture to production deployment, I deliver measurable results.
About Me
I'm a Machine Learning Engineer and Data Scientist with 2 years of experience building AI-driven solutions that go beyond proof-of-concept. My focus is on production-grade systems: models that run reliably in real-world environments, handle scale, and deliver business value.
🎓 Education
Master's Degree in Computer Science
Polytech Marseille
Artificial Intelligence, Machine Learning, Data Science
📜 Certifications
MLOps Practitioner & Advanced Designer
Dataiku • 2024
Google Cloud Professional Data Scientist
Google Cloud • In Progress
End-to-end ownership
I don't just train models. I architect data pipelines, containerize services, set up monitoring, and deploy to production. My projects span the full ML lifecycle.
Production-first mindset
Whether it's a real-time trading bot processing Kafka streams or an automated code review system integrated with GitHub, my work is designed for reliability, not just demos.
NLP/LLM expertise
I've fine-tuned transformer models for financial sentiment analysis, built RAG systems with vector databases, and integrated LLMs for summarization and generation tasks.
Pragmatic problem-solving
I choose technologies based on requirements, not hype. Sometimes that means FAISS over a managed vector DB, or a well-structured FastAPI service over a heavyweight framework.
Professional Experience
Bouygues Telecom
Data Scientist (Alternance)
March 2024 - March 2025
Meudon, France
- ▶Collaborated with business teams to identify criteria for prioritizing high-potential calls
- ▶Designed a scoring model (XGBoost) prioritizing high-conversion potential calls: +46% useful contact rate, 94% accuracy
- ▶Built complete CI/CD pipeline with automated tests (accuracy, latency), progressive deployment and drift monitoring
- ▶Reduced time-to-market by 30% through deployment automation (Docker, Azure ML)
LeBonCoin
Data Scientist (Internship)
April 2023 - October 2023
Paris, France
- ▶Automated classification of user ads (manual categorization was causing delays and errors)
- ▶Built NLP model combining TF-IDF and embeddings + Random Forest: accuracy from 78% to 92% across 15 categories
- ▶Deployed to production via FastAPI microservice on Vertex AI, real-time processing of new ads
Universidad Politécnica de Valencia
Data Scientist (Internship)
December 2022 - March 2023
Spain
- ▶Built MAGENTA dataset: 25K+ Arabic texts for AI-generated text detection (3 domains)
- ▶Generated synthetic texts via LLM (AraGPT-2, BLOOM) and fine-tuned AraBERT: 96.4% Macro-F1
- ▶Publication: ICALP 2023 (International Conference on Arabic Language Processing), Springer
Marsa Maroc
Data Analyst (Internship)
July 2022 - August 2022
Mohammadia, Morocco
- ▶Analyzed maintenance data (sensors, ERP, intervention history) for transition from curative to predictive maintenance
- ▶Built ETL pipeline to centralize heterogeneous sources in Snowflake and created Tableau dashboards (-40% reporting time)
Featured Projects
FinSentBot
Real-Time Trading Signal Generation
Built an end-to-end automated trading intelligence system that combines real-time news sentiment analysis with live market data to generate Buy/Hold/Sell signals.
🎯 Problem
Financial traders need to process vast amounts of news and market data to make informed decisions. Manual sentiment analysis is slow, subjective, and can't keep pace with market movements.
💡 Solution
Developed a production-ready system using Apache Kafka for real-time data streaming, FinBERT for financial sentiment analysis, and custom ML models for signal generation.
📊 Key Results
- ✓Automated analysis of 100+ financial news articles daily
- ✓87% accuracy on financial sentiment classification
- ✓End-to-end latency under 5 seconds from news to signal
- ✓Production-ready with comprehensive logging and error handling
Code Review AI
Automated Python Analysis System
Developed an AI-powered code review system that combines Abstract Syntax Tree (AST) parsing with Claude AI to automatically analyze pull requests and provide detailed, actionable feedback.
🎯 Problem
Manual code reviews are time-consuming and often miss subtle bugs, security issues, or style inconsistencies. Development teams need automated quality checks that integrate seamlessly into their workflow.
💡 Solution
Created a production system using FastAPI webhooks, Python AST parsing, and Claude AI for intelligent code analysis with automatic PR commenting.
📊 Key Results
- ✓Deployed to production on Railway with live webhook integration
- ✓2,400+ lines of production code with modular architecture
- ✓Average review time reduced to <30 seconds per PR
- ✓100% automation with zero manual intervention required
DocuMind
Intelligent RAG System for Semantic Search
Built a Retrieval-Augmented Generation (RAG) system that combines semantic embeddings with vector similarity search to enable intelligent document discovery.
🎯 Problem
Traditional keyword search fails to capture semantic meaning, making it difficult to find relevant information in large document collections. Users need intelligent systems that understand context and intent.
💡 Solution
Implemented a RAG architecture using Sentence Transformers for embeddings, FAISS for vector search, and planned LLM integration for generation.
📊 Key Results
- ✓123 documents indexed with sub-second query response times
- ✓92% of test queries return relevant results in top-3
- ✓Fully containerized MVP ready for production deployment
- ✓Roadmap includes LoRA fine-tuning and multi-modal support
AI News Agent
Automated Multi-Source News Pipeline
Created an end-to-end automated pipeline that scrapes major tech publications, deduplicates and processes articles, generates AI summaries, and distributes daily email digests.
🎯 Problem
Staying current with technology news across multiple publications is time-consuming. Professionals need curated, summarized content delivered automatically without manual aggregation.
💡 Solution
Built a fully automated pipeline using Selenium for scraping, OpenAI for summarization, and GitHub Actions for scheduling.
📊 Key Results
- ✓Automated aggregation from 5 major tech publications daily
- ✓100% automated pipeline with zero manual intervention
- ✓French summaries optimized for quick consumption (2-3 sentences per article)
- ✓Free hosting via GitHub Actions with no server costs
- ✓Average 20-30 articles processed and summarized daily
Technical Skills
Programming & Data
Machine Learning & AI
NLP & LLMs
MLOps & Deployment
Cloud & Infrastructure
Data Visualization & Tools
Let's Connect
I'm currently available for full-time ML Engineer / Data Scientist roles and freelance AI/ML projects. Whether you're looking to build production NLP systems, deploy real-time ML pipelines, or implement RAG solutions, I'd love to hear from you.
📍 Based in Paris, France
Open to remote opportunities across Europe and international projects
⏱️ Response Time
I typically respond within 24 hours.