36: ML Ops: Data Models into Production | Adria Salvador, Glovo

“Only 22 percent of companies using machine learning have successfully deployed a model”.(deeplearning.ai ) Why such a big gap between machine learning model development and production? What are the big challenges and how are they being solved? Adrià Salvador, lead of the Data Science Productization team at Glovo, is here with us today to share his tricks and tips in the field of ML Ops. From his honest revelation of personal experiences, we see the challenges leading a team to productionize data science directly into the company’s operation.  Other than real case references,  Adrià also introduced concepts of best data science practices, machine learning model production pipeline in Glovo, and open source tool packs for us to add to our next learning list. Don’t miss the opportunity and listen in! Speaker Bio: Adrià Salvador Palau (Barcelona, 1990) holds a BSc and MSc in Physics. He also holds a PhD in Engineering from the University of Cambridge.  In his PhD, Adrià developed distributed machine learning architectures to predict failures in large fleets of industrial machines. Adrià’s research focused both on the technological and economical challenges of implementing these technologies in industrial scenarios. He joined Glovo two years ago to work as a Data Scientist. Since then, he has been promoted to lead the Data Science Productivisation team at Glovo. His team has the responsibility of speeding up productivisation of machine learning models in glovo and helping determining MLOPS best practices within the company. Resources: MLOPS by google MLOPS in Towards Data Science Deep Learning by Goodfellow Umap Paperswithcode Poetry