Tanguy Esteoule

Formerly Trailstone & Meteo*Swift · retained through two consecutive acquisitions

1. Senior Machine Learning Engineer

   2024 – present
   • Designed and scaled a fully serverless forecasting platform (AWS Lambda, Step Functions) processing real-time data for 1,000+ renewable assets with <15-min latency.
   • Raised availability to 99%+ with proactive observability (CloudWatch, automated alerting), reducing mean time to recovery.
   • Lead applied research on forecasting accuracy and turn research prototypes into scalable, production-ready ML pipelines.
   • Drove the adoption of LLM-based coding agents: agentic workflows are now fully part of how I design, build and operate our systems.

2. Machine Learning Engineer

   2022 – 2024
   • Contributed to the migration from a monolith to serverless AWS microservices, cutting cloud costs by 20%.
   • Implemented Terraform IaC and Bitbucket CI/CD pipelines, bringing deployments from days to minutes.
   • Defined a standardized path-to-production for ML research (tests, versioning, packaging) adopted by the data science team.

3. Data Scientist

   2016 – 2022
   • Developed wind power forecasting models based on machine learning rather than physical models, outperforming industry benchmarks.
   • Built early MLOps foundations in a startup environment: MLflow, experiment tracking, reproducible research workflows.
   • Started as a CIFRE industrial PhD, then brought the thesis research into the wind farms managed by the company.

Adaptive Multi-Agent Systems for Wind Power Forecasting

• Designed an Adaptive Multi-Agent System (AMAS) where cooperating agents forecast wind power, learning from imperfect numerical weather predictions.
• Detected and accounted for wake-effect interactions between wind turbines.
• Worked on ensemble weather forecasts with the French national weather research centre (CNRM).