Pulpix - Video Recommendation at Scale
- 1. Video Recommendation at Scale January 2017 RecSysFR Meetup
- 4. About Pulpix
- 5. About Pulpix AI startup based in New-York and Paris 100+ sites using our technology in the world $850,000 seed round raised in the US Accelerated in Silicon Valley by
- 6. About Pulpix AI startup based in New-York and Paris 100+ sites using our technology in the world $850,000 seed round raised in the US Accelerated in Silicon Valley by
- 7. About Pulpix 100+ sites using our technology in the world Accelerated in Silicon Valley by $850,000 seed round raised in the US AI startup based in New-York and Paris
- 8. About Pulpix 100+ sites using our technology in the world $850,000 seed round raised in the US AI startup based in New-York and Paris Accelerated in Silicon Valley by
- 9. What do we do at Pulpix?
- 11. How?
- 12. People binge-watch videos on social platforms
- 13. Why?
- 14. UX AI There is one simple reason: they invest in UX and AI.
- 15. Media Website vs. Social Platform Media Website Media Website
- 16. So, what do we do at Pulpix?
- 17. 1 7
- 18. 1 8
- 19. 1 9
- 20. 2 0
- 21. 2 1
- 22. 2 2
- 23. 2 3
- 24. 2 4
- 25. 2 5
- 26. 26
- 27. 27
- 28. 28
- 29. 29
- 30. 30
- 31. 31
- 32. 32
- 33. 33
- 34. 34
- 35. 35
- 36. 36
- 38. Video to Video
- 39. Video to Video
- 40. Article to Video
- 41. in Video
- 43. Recommendation at Scale Key figures 10 million videos Less than 100 ms response time 10 million events a day More than one billion training events
- 44. Recommendation at Scale Key figures 10 million videos Less than 100 ms response time 10 million events a day More than one billion events
- 45. Less than 100 ms response time More than one billion training events 10 million events a day 10 million videos Recommendation at Scale Key figures
- 46. Less than 100 ms response time 10 million events a day 10 million videos More than one billion training events Recommendation at Scale Key figures
- 47. From R&D to production Iterate fast
- 48. From R&D to production Iterate fast Idea Prototype Retrospective evaluation
- 49. From R&D to production Iterate fast Idea Prototype Retrospective evaluation A/B test 24h to 10 days
- 51. Content-based Engine First approach Speech-to-text Metadata Keywords extraction Weighting
- 52. Content-based Engine First approach Speech-to-text Content ScoreMetadata Keywords extraction Weighting
- 53. Content-based Engine First approach Speech-to-text Content ScoreMetadata Keywords extraction Weighting Recency boost
- 59. Example L’Equipe.fr Football Basket Tennis 54% 12% 26% User preferences
- 60. Collaborative Filtering How to put it into practice? • User-based recommendations - Known users only - Not contextual
- 61. Collaborative Filtering How to put it into practice? • User-based recommendations - Known users only - Not contextual • Video-based recommendations - For all users - Fully contextual
- 62. Hybrid Engine
- 66. Personalization RecencyContext Hybrid engine Key dimensions
- 67. Hybrid Engine Linear score combination Content Engine Collaborative Engine Context Recency Global score Scores Linear Model
- 68. Hybrid Engine Nonlinear embedding combination Content Engine Collaborative Engine Context Recency Scores Candidate videos
- 69. Hybrid Engine Nonlinear embedding combination Content Engine Collaborative Engine Context Recency Global score Scores Candidate videos Nonlinear Model Features
- 70. Scalable Data Science stack
- 71. Flexibility queuing Scalability sharding Fault tolerance replication High throughput replication Recommendation at Scale Requirements
- 72. Flexibility queuing Scalability sharding Fault tolerance replication High throughput replication Recommendation at Scale Requirements
- 73. Flexibility queuing Scalability sharding Fault tolerance replication High throughput replication Recommendation at Scale Requirements
- 74. Flexibility queuing Scalability sharding Fault tolerance replication High throughput replication Recommendation at Scale Requirements
- 75. Recommendation at Scale Our Data Science stack
- 76. Recommendation at Scale Our Data Science stack tracking
- 77. Recommendation at Scale Our Data Science stack tracking ML data processing
- 78. Recommendation at Scale Our Data Science stack tracking reco ML data processing
- 79. What’s next?
- 80. • Reinforcement Learning What’s next? Our current R&D
- 81. • Reinforcement Learning • Deep Learning: ○ Recommendation ○ Video recognition What’s next? Our current R&D
- 82. @pulpix denis.vilar@pulpix.com PARIS NEW-YORK sabry.otmani@pulpix.com 124 rue d’Aboukir 75002 Paris, France 584 Broadway New York 10012 NY, USA Pulpix Pulpix Inc. hello@pulpix.com +33 (0)6 66 15 02 42 +1 (415) 996 4453 www.pulpix.com Pulpix is recruiting! jobs@pulpix.com