Senior Data Engineer

Our client is a legal tech startup that focuses on AI and machine learning, specifically building chatbots to answer legal questions for lawyers. They are looking for a Senior Data Engineer for a high-impact project: digitizing law in Morocco and Africa and creating the first AI-quarriable legal knowledge base.

Their ambition is to build a platform capable of answering legal questions in a reliable, well-sourced, and traceable way, based on a massive corpus of heterogeneous legal documents.

🚀 Why this project is different
You will join a true “knowledge infrastructure” mission:

• Contribute to making the law more accessible
• Build a durable asset: a structured database of Moroccan law (in French), extensible to Africa
• Work on a concrete and deep technical challenge: transforming unstructured data into exploitable, reliable, and maintainable data at scale

Required skills:

• 3+ years of experience in Data Engineering and/or applied Document AI / NLP
• Strong proficiency in Python
• Hands-on experience with unstructured documents: PDF parsing, OCR, cleaning, structuring
• Used to delivering to production: robust pipelines, observability, quality, performance
   

🛠 Stack/skills (indicative)

• Storage: AWS
• Document processing: OCR/parsing tools, text preprocessing pipelines
• Testing & quality: metrics, sampling, automated validatio

⭐ Nice to have

• Experience with legal / regulatory corpora or high-precision content
• Familiarity with multilingual issues and encoding
• Basic knowledge of downstream needs (vector DBs, retrieval, citation)

Scope of work:

You will be responsible for the “documents → structured data” pipeline that will feed our AI (RAG) engine.

At the core of the role (technical focus)
Build a structured database of Moroccan law in French from highly heterogeneous data:

• PDFs (text-based and scanned), Word files, images, text files, sometimes noisy or incomplete
• Text extraction (parsing + OCR when needed), cleaning
• Structuring: detection of titles/chapters/sections/articles, hierarchy, normalization
• Intelligent chunking (based on legal structure rather than arbitrary size), with traceability (source, page, identifiers)
• Metadata: date, type of text (law/decree/circular/case law, etc.), source, version, article numbers, etc.
• Deduplication & versioning: redundant documents, amendments, consolidated versions
• Industrialization: orchestration, logs, retries, idempotence, monitoring, quality tests