Looking for a job

Amar Merabti

@amarmerabti

Message

Data Scientist at Nricher who built product data quality control clearing 68% of values at 99% precision.

France

Message

What I'm looking for

I'm seeking a permanent position (CDI) and am available immediately.

At Nricher, I built an automatic quality-control system for product attributes that cleared 68% of extracted values without human review at 99% precision. I also created a reproducible pipeline for a labelled dataset and optimised GPU inference, achieving a 20 to 50× speed-up.

At ENAGEO (Sonatrach group), I worked on machine-learning models to reconstruct missing well measurements and predict reservoir-rock properties. My projects include a Kabyle language model trained from scratch and a real-time e-commerce recommendation engine using Kafka and PySpark.

Experience

Work history, roles, and key accomplishments

Data Scientist (final-year internship)

Nricher

Feb 2026 - Aug 2026 (6 months)

Designed an automatic quality-control system for extracted product attributes using mDeBERTa-v3 and TabPFN, achieving 68% of values cleared without human review at 99% precision. Optimized GPU inference with batched forward passes and fp16 autocast, achieving 20-50x speed-up.

Python PyTorch DeBERTa V3 TabPFN NLP GPU Inference Pandas Data Pipelines

AI Developer (Machine Learning)

ENAGEO

Jul 2023 - Jul 2024 (1 year)

Reconstructed missing well measurements using LSTM and RNN models, achieving MAE 12.7 and RMSE 15.2. Improved cross-well generalization with correlation increasing from 0.52 to 0.82.

Python LSTM RNN Time Series Imputation

Machine Learning Intern

ENAGEO

Feb 2023 - Jun 2023 (4 months)

Predicted geomechanical properties of reservoir rock using linear regression, RBF-SVM, and MLP, reducing MAE from 27.4 to 21.4. Delivered an application for CSV import/export, model training, and user management.

Python scikit learn Java MySQL UML