Why ML in Production is Still Broken

Explore the challenges of implementing machine learning in production environments in this 44-minute conference talk by Hamza Tahir, CTO at maiot GmbH, presented at the Toronto Machine Learning Series (TMLS). Discover why 87% of machine learning projects fail to reach production and the disconnect between development in Jupyter notebooks and real-world application. Examine the key differences between machine learning and traditional software engineering, and learn why treating data as a first-class citizen is crucial for successful production ML systems. Gain insights into the ongoing struggle to meet quality standards in ML production, despite the circulation of the influential Hidden Technical Debt paper since 2017.

Hamza Tahir - Why ML in production is STILL broken?