Freelancer profile translated to English.
Back to original languageAbout Siwar
PhD in Artificial Intelligence, I have been designing and deploying high-impact NLP and Generative AI systems for over 12 years, from academic research to Python production.
My core business: transforming unstructured data (texts, documents, emails, contracts, financial reports) into actionable insights using LLM, RAG, and deep learning architectures.
What I bring concretely to your projects:
- LLMs & RAG: design of on-premise (Llama 3) or cloud (GPT-4) pipelines with semantic chunking, re-ranking, and rigorous evaluation, up to 70% measured productivity gain.
- Advanced NLP: multi-label classification, NER, information extraction, embeddings, fine-tuning of pre-trained models on your business data.
- Document Intelligence: automatic extraction of financial tables (96% accuracy), PDF processing and OCR, end-to-end pipeline.
- Python Production: maintainable codebase, FastAPI, CI/CD, cloud deployment (AWS, Azure) or on-premise for confidentiality constraints.
- Applied research: 23 peer-reviewed publications (200+ citations, h-index 8), supervision of engineers and PhD students, benchmarking of SOTA models.
I work both autonomously on a specific subject and as a reinforcement to an existing Data/AI team. Available for consulting, prototyping, POC, or long-term development missions.
Based in Paris · Remote possible · Available quickly
French
Native or bilingual
English
Fluent
Arabic
Native or bilingual
Can work on-site
Paris (up to 50km)
Experience
- Upskills R&DSenior data scientist R&DBANKING AND INSURANCEAugust 2019 - June 2025 (5 years and 10 months)Paris, FranceDesign and deployment of NLP and Generative AI systems in production for major clients (banking, finance, trading). End-to-end responsibility: architecture, experimentation, evaluation, delivery.Key projects:
- On-premise RAG (Llama 3) for financial contract analysis: -70% processing time, 92% accuracy, 100% data privacy.
- Multi-label email classification (DBS Banking): F1 from 70% → 90%+ after data quality diagnosis + deep learning fine-tuning.
- Automatic PDF financial table extraction (YOLOv3 + NLP): 96% detection, published DEXA 2022.
- FuzzyART clustering on trading data: 96% dataset reduction, 97% of root cause information preserved.
- Upgrad (Université en ligne),Thesis SupervisorEDUCATION AND E-LEARNINGJune 2025 - Today (1 year)Supervise and mentor PhD students in business analysis to help them successfully complete their thesis projects on topics related to Gen AI and innovation, with various applications such as HR, e-commerce, and other emerging fields.
- Université Savoie Mont BlancPOSTDOC MULTIMODAL AIRESEARCHDecember 2017 - May 2019 (1 year and 5 months)Annecy, FranceResearch in multimodal recognition (vision + classification) with application on real data. Design of an offline mobile system optimized for embedded constraints. Publication of several scientific articles.Development of a multimodal tree species recognition system (leaf + bark) via Evidence Theory + EkNN; +19 pts accuracy vs state-of-the-art (56% → 75%).
Recommendations
Be the first to recommend Siwar
Help this freelancer shine by sharing your experience working together.
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- PhD in Artificial IntelligenceUniversité de Tunis & Université de Rennes 12016Doctorat en intelligence artificielle
- Master in Computer Science applied to ManagementInstitut supérieur de gestion de Tunis2012Master en Informatique appliquée à la gestion