Meetup 3: Machine Learning Reproducibility, Experiments, and Pipelines Automation for NLP

14 October
2019, Amsterdam

First ML REPA meetup in Amsterdam! This is a practical meetup for machine learning (ML) practitioners, team leads, data scientists and ML projects managers. Benchmark of approaches to NLP pipelines automation: AI plays words games
18:00 - 18:15
18:00 - 18:15
Registration, pizza, beers, networking
18:15 - 18:45
18:15 - 18:45
ML experiments management, pipelines automation and reproducibility: with DVC and MLFlow
It starts with a review of different technical approaches to organize the work on an example ML task. From all-in-one Jupyter Notebook to python scripts and automated pipelines.
18:45 - 19:05
18:45 - 19:05
ML teams and projects management: potential for cost optimization
Then it explores ML project workflow and managing experiments requirements to highlight major issues and potential for time cost optimization.
What problems we encounter in an attempt to automate ML? How we share best practices in a large company with distributed ML teams? How well are these approaches applicable in different ML projects for Digital Marketing, HR, Text Analytics, LifeTime Value predictions.
19:05 - 19:15
19:05 - 19:15
19:15 - 20:00
19:15 - 20:00
Benchmark of approaches to NLP pipelines automation: AI plays words games
Finally, it covers an NLP case with examples of code organization and experimentation pipelines with different open source tools like DVC, MLflow, kubeflow, etc.
Achieve faster experiments, easy collaboration and reproducible results with one-magic-button to run end-to-end experiments.
Mikhail Rozhkov
ML REPA founder & Open Data Science (ODS Russia) active member
Senior Data Scientist, Raiffeisenbank Russia
I start to learn and practice Machine Learning few years ago. Work on different ML applications for search engine optimization, HR and Digital Marketing. Worked with NLP, CV, time series and tabled data.

Now I focus on optimization of Data Science teams collaboration and Machine Learning automation. There is a huge potential for ML tasks automation, do better experiments and model management. With ML REPA we learn, apply and share tools for models and data version control, reproducible experiments management and automated pipelines.
Nieuwendammerkade 26A-5 ยท Amsterdam

Our partners
email: info@ml-repa.ru
Apply to speak!
Made on