Machine Learning in our apps

Transcript

1. Machine Learning in our apps NOVEMBER 2019
2. Systems Engineer Universidad Libre de Colombia Whoami Analytics and Machine Learning Pontificia Universidad Javeriana Engineering Manager Merlin Jobs Co-Organizer IOS Developers Colombia
3. “The field of study that gives computers the ability to learn without being explicitly programmed” Arthur Samuel Learn = Find complex patterns in the data. Learn algorithms Automatic = Systems will have continuous improvements through learning Machine Learning
4. Garbage in Garbage out
5. Reinforcement Learning Unsupervised Learning Supervised Learning Machine Learning
6. Machine Learning Process ML Models Training Inference or prediction Clean Data
7. Is possible to do any ML stuff in mobile? NO Training – Metal has some APIS for on Device training and CoreML 3 Faster on prediction tasks User privacy If we choose the right use case, it could be cheaper
8. Tools
9. Text Classification Train/ Test Economy Sports Technology
10. CreateML App
11. CreateML Code
12. CreateML Code
13. Machine Learning at Merlin Jobs Ranking Matching Jobs/Candidates
14. Ranking System Merlin System Ranking System using Score
15. Thinking in our mobile model Merlin System Ranking System using Score – Heuristics Model User complaints High Latency – More than 3 secs Ranking is only used in the user experience “Faster response times: It’s more responsive than sending HTTP requests, so doing real-time inference is possible, making for a better user experience” 300 Milliseconds
16. Feature Engineering 1.Requirements 2.Distance 3.Profile data 4.Matching in positions levels 5.Score
17. Our Mobile Model
18. Results version 0.0001
19. Results version 0.000001 Merlin System Ranking System using Score – Heuristics Model Happy users Reduce Latency Predictive model to determine good and bad match Prediction 1000 tuples = 89 Milliseconds
20. Concluding 1. Faster Mobile Inference 2. You don’t need to be an expert in ML 3. The most important thing: feature engineering and data cleaning 4. Improve user experience 5. Find use cases in your companies to start using ML in our apps
21. THANK YOU!