Real Data Science Examples at OLX's Online Marketplace
summary
OLX runs an automatic moderation pipeline that saves uploads to S3 → hashes → Elasticsearch, computes image and metadata features, and runs a duplicate-detection ML model that either auto-rejects or sends uncertain cases to human moderators and then uses that feedback to retrain. Recommendations use collaborative filtering with implicit (ALS) + neural embeddings to produce 16–32 dimensional vectors for personalized home feeds and item-to-item suggestions, deployed as microservices (Flask/AWS) to pre-moderate and scale.
AI-written summary · about
chapters (8)
- 0:00Introduction and welcome
- 2:03What is OLX marketplace
- 5:15Data science areas at OLX
- 12:19Content moderation system
- 18:39Recommender systems overview
- 25:02Data science team collaboration
- 29:54Expectations from data scientists
- 34:06Q&A session
related talks
48mFrom (Big) Data Mess to Data as an Innovation Enabler
33mMachine Learning to moderate ads
44mRobust AI Search Ranking for Radical C2C Marketplace Growth
23mFrom Data to Recommendations: Building an Intelligent System with ROR
29mBuilding ML Platform Capabilities at Global Scale
30mEvolving Search in Online Marketplaces: Practical Applications of AI
39mDeveloping Machine Learning for Impact
23mDeep Recommendations for e-Commerce