In this presentation, Ankit Raheja, helps you understand whether it makes sense to build AI Products and how to showcase the value you can get out of your AI Products. He also discusses what you should focus on during Designing Products. And finally, talks about how Developing and Deploying AI Products are two very different beasts and how to deal with them differently.
16. What should an AI Product Manager Do
1. Come up with a Solid Business Model for Path to Profit
2. Understand whether it an AI solvable problem
a. Too much Data - Google Search
b. Computer Vision - Amazon X Ray Feature
c. NLP - Siri/Alexa/Google Assistant
d. Anomaly Detection - Payment Providers
e. Recommendations
f. Reinforcement learning
g. Predictions and Forecasting
h. Others
3. Strategic Data Acquisition - Training/Feedback Data is the new wireframe
4. Understand ML concepts and define & track success metrics*
5. PM 101 Skills
6. Agile* Data Science
18. Why is DL picking up and why should you care?
1. Data Explosion 4. Robust,Generalizable and Scalable
Algorithms
2. Computation Power - CUDA 5. Neural Networks are cool again
3. Less time feature engineering 6. Activation Functions*
3584!
24. Dropout Rate
You drop off neurons to generalize better and to avoid overfitting and memorizing
https://codelabs.developers.google.com/codelabs/cloud-tensorflow-mnist/#9
28. Machine Learning in Production
● Choose Right models for Offline or Online Processing
● Response Time, Implementation Cost, Runtime Cost
● Choose Containers for Productionizing - Packaging, Inheritance,Microservices
● Track Metrics in Real Time
● Put Fail Safe Mechanism in place
● Get Evangelists
● Training Data/Test Data is indicative of Real Data
● Reproducible Research Tools/Algo Data Banks
● Run Offline Models first before putting them in production
● Run Targeted A/B Tests ensuring proper Statistical Significance, Sample Size etc.
● Model Development Lifecycle ≠ Software Development Life Cycle
29. Next Steps
Great Resources
1. Prediction Machines - Simple Economics of AI
2. Follow/Youtube - Geoffrey Hinton , Andrew Ng, Yann LeCun,Andrej
Karpathy,Ian Goodfellow,Fei-Fei Li etc.
3. DeepLearning.AI
4. Fast.AI
5. Hands-On Machine Learning
6. Deep Learning Fundamental: An Introduction for Beginners
7. Siraj Raval’s videos
8. https://keras.io/
31. www.productschool.com
Part-time Product Management, Coding, Data, Digital
Marketing and Blockchain courses in San Francisco, Silicon
Valley, New York, Santa Monica, Los Angeles, Austin, Boston,
Boulder, Chicago, Denver, Orange County, Seattle, Bellevue,
Toronto, London and Online
Editor's Notes
Training Time from Weeks to Days. Can Iterate Quickly. Data is exploding. So much data is getting generated and easier to ingest now. Parallel Computing Architecture. 3584 cores. Throw Kitchen sink at it. Tune hyper parameters
Input layer or Input Features , Hidden Layers - Pick 4-5-6-7-8,