Data Scientist.Employment

Raw agricultural signals (weather, crops, soil, yields) arrived in heterogeneous formats with no unified storage or compute layer, blocking the team from turning the data into product features.

Designed and deployed a big data architecture combining a Hadoop data lake, an Apache Spark cluster, and ElasticSearch. Ingested 100GB+ of agricultural data from multiple sources (scraped market data, spreadsheets, PDFs, images) covering crop information, fertilizers, yield production, rainfall, and climate. Built end-to-end ingestion and transformation pipelines turning the raw data into ML-ready features for the smart-farming app.

Business strategy on planting cycles, sourcing and pricing was made without a quantitative view of expected yields, leading to costly reactive decisions.

Developed statistical crop yield forecast models using regression algorithms, fed by the agricultural data pipelines, mainly for maize with additional models for beans. Achieved R² above 80% in most cases. Translated forecasts into a predictive layer for business planning: sourcing, cycles, and pricing decisions.

Leadership needed visibility into the analytical insights, but raw outputs were inaccessible to non-technical readers.

Delivered a consolidated leadership BI report in Kibana and Apache Superset, covering multiple KPIs and data domains.

Société Générale Maroc’s top management needed structured exposure to data and AI applications in banking to inform strategic priorities.

Designed training content and developed lab use cases tailored to banking applications of data and AI, delivered to Société Générale Maroc’s top management in a single training mission.