Machine Learning for Developer Self-Care

" summary

Meta handles 2.89 billion monthly users and 4 petabytes of data per day and trains models that turn that data into developer tools. Built systems include neural code search using vector embeddings, Aroma code-snippet search and autocomplete, Getafix automatic patching that uses anti-unification to learn fixes from commit history, ranking for predictive test selection, simulated “WW” digital twins for outages, and large code models for review and ownership inference — all de‑risked and running with product teams.

AI-written summary · about

" related talks

• 43mMachine Learning for Developer ProductivitySatish Chandra · Strange Loop 2022
• 37mMachine Learning on Source CodeFrancesc Campoy Flores · Scale By The Bay 2018
• 24mMachine Intelligence gets realBabak Hodjat, Karin Klein, Todd Levy, Drew Conway · DLD New York 2017
• 33mMachine Learning 2.0 with Hugging FaceJulien Simon · Conf42 Machine Learning 2022
• 54mMachine Learning for DevelopersDanilo Poccia · Voxxed Days Zurich 2017
• 12mUsing GenAI to Accelerate Privacy ImplementationsRituraj Kirti · PEPR '25
• 57mBuilding smarter solutions with Machine Learning, from magic to realityLaurent Picard · code.talks commerce 2019
• 46mThe AI-Powered DeveloperJohn Fletcher, Daniel Hartung · DevDays Europe 2024