MSc Data Sciences & Business Analytics · CentraleSupélec / ESSEC · Paris

I am a data scientist and analyst with a background that spans early-stage startups, agri-tech commercial roles, and academic research in deep learning. I started my career in sales and key account management across Dubai and India, where I developed a sharp instinct for operational problems and customer needs. That foundation led me toward data: first building reporting systems and ETL pipelines at Nila Cares, then deepening my technical skills through an MSc at CentraleSupélec and ESSEC, where I am currently completing a deep learning research project with SLB on well-log formation classification.

I work across the full data stack: from pipeline architecture and SQL modelling to computer vision, LLM-powered applications, and analytics dashboards. I am most energized by problems where data engineering and real-world impact meet directly.

Currently seeking internship opportunities in data engineering, solutions engineering, and analytics. Based in Paris. Open to hybrid and remote roles.

Languages: Python · SQL

Data & Analytics: Pandas · NumPy · PostgreSQL · Metabase · NocoDB

Machine Learning: PyTorch · Scikit-learn · YOLO · ResNet18

AI / LLM: OpenAI API · RAG pipelines · Claude API · DistilBERT (LoRA fine-tuning)

Infrastructure: ETL pipelines · Docker (learning) · GCP · Git

Viz & BI: Metabase · Power BI · Matplotlib

MSc Capstone · CentraleSupélec / ESSEC in partnership with SLB · 2025 – 2026

A deep learning research project tackling automated geological formation classification from well-log data, in collaboration with SLB's data science team. The project addresses a core challenge in subsurface analytics: classifying formations across the North Sea (Hordaland, Rogaland, Shetland, Viking, Vestland) and Colorado datasets reliably and at scale, reducing dependence on manual expert labeling.

Supervised Segmentation Pipeline

• Adapted YOLO11n with Focal Loss for well-log image segmentation, treating formation boundaries as a dense prediction problem across multi-channel log inputs.
• Trained and evaluated across North Sea and Colorado geological datasets, optimizing for class imbalance using Focal Loss to handle rare formation types.
• Built explainability layer using Guided Grad-CAM to surface which log features drive formation predictions, enabling geologist review and trust calibration.

Hybrid Time-Series Architectures

• Designed and benchmarked LSTM-FCN-2dCNN and LSTM-XCM hybrid architectures for sequential well-log classification, combining temporal modelling with convolutional feature extraction.
• Evaluated architectures across cross-well generalization tasks to assess robustness to geological variability between datasets.

VLM-Based One-Shot Transfer Pipeline

• Developed a vision-language model pipeline enabling one-shot transfer of formation classifications to new wells with minimal labeled data.
• Designed prompt logic and few-shot inference strategies to leverage VLM spatial reasoning for geology-specific pattern recognition.

Infrastructure

• Deployed training workloads on a GCP VM (N1, Tesla T4, 120 GB SSD) to accelerate deep learning experiments.
• Managed experiment tracking, versioning, and collaborative development via Git across a team of five researchers.

Team: Enora Barbier · Yann Djouka · Alessandro Ivashkevich · Faidon Kotsakis
SLB Contacts: Shwetha Salimath · Sylvain Wlodarczyk

Data Analyst Intern · Nila Cares — Migrated data architecture from Airtable to NocoDB, built Python ETL pipelines, and engineered Metabase dashboards used for city expansion decisions.

Key Account Manager · Waycool Foods — Grew agri-export revenue by 350% to $500k/month, built Excel-based live margin tracking system.

Sales Development Associate · FarmUnboxed — Data-driven sales forecasting that reduced stockouts by 30%.

🎓 MSc Data Sciences & Business Analytics — CentraleSupélec / ESSEC (2024 – 2026)
🎓 BA Economics, First Class Honours — University of Manchester (2016 – 2019)

📧 george.emmanuel@essec.edu
💼 LinkedIn