Developing Machine Learning for Impact
summary
A machine learning product manager explains concrete practices at InfoJobs / Adevinta: building recommendation systems that become one of the main sources of applications, a CV parsing pipeline that shortens signup and yields more complete profiles, and a normalizations taxonomy that maps synonyms (programmer ↔ software developer) to improve matching. Teams run A/B experiments, deploy models as APIs via ML engineers, monitor data quality, accept high uncertainty in feasibility, and prefer small MVPs before adding ML.
AI-written summary · about
chapters (10)
- 0:00Intro
- 2:00Why this talk?
- 5:02ML as a tool, not a solution
- 8:40Less is often more
- 17:36Without data quality, there is no ML
- 19:58Surviving the ML uncertainty
- 23:40The importance of multidisciplinary teams
- 27:47What is not measured can't be improved
- 31:40Success stories
- 38:24Outro
related talks
49mThe Impact of Machine Learning in Companies
22mBefore the Model: How Machine Learning Products Start...
27mA Look Inside Applied Machine Learning
24mMachine Intelligence gets real
22mReal-world applications of machine learning: Beyond the hype
15mMachine Learning for All
33mAI with a DevOps mindset - experimentation, sharing ..
01h 30mBecome a data-driven Organization with Machine Learning