AI for Product Managers & Data Leaders

Vous aurez acquis les connaissances suivantes :

• Savoir définir ce qu’est l’Intelligence Artificielle et le Machine Learning
• Savoir juger lorsque le Machine Learning est nécessaire dans un projet data et quand il ne l’est pas
• Comprendre le lifecycle de tout projet à base de Machine Learning
• Comprendre les spécificités inhérentes à la Data
• Appréhender les étapes nécessaires à la création de Produits à base de Machine Learning

‍

Product Owners, Product Managers, Data Leaders

Des notions de Product Management sont un plus.

Part 1: The basics of Machine Learning

Theory

• What is ML
• Importance of data: garbage in, garbage out
• Supervised vs. Unsupervised
• Tabular vs. Unstructured data (computer vision & NLP)

Good use cases for ML

• Examples from automation
• Examples from optimization

Bad use cases for ML

Part 2: Introduction, Role of AI in product management

Product Discovery

• Identifying and assessing opportunities with data and AI
• Mitigating risks: Value, Usability, Feasibility, Business Viability

Product Planning

• Including AI in product vision
• Roadmap planning: managing high uncertainty
• what metrics to use

Product Development

• Data-First mindset
• Experiment-driven development
• Think of the MVP as the starting point for further experimentation

Product Maintenance

• Challenges of ML in production
• Optimization
• Model / Data drift

Product retirement and when to stop an AI initiative

Part 3: What are the steps to develop a ML product?

Stages of a typical ML project cycle

• Collecting and preparing the data
• Building and evaluating the model
• Model deployment
• People who may be involved

Data

• Data Quality
• What are the common issues with data
• Data annotation: how to approach

Models and algorithms

• Recommender system
• Classifier
• Deep learning / neural network

Technology

• Get familiarized with some common tools
• Examples : Scikit Learn, PyTorch, TensorFlow, GPUs

CRISP-ML(Q) framework

CRoss Industry Standard Process for Machine Learning with Quality Assurance

1. Business understanding (from business objectives to data science goals)
2. Data understanding (what data you have, what data you need, etc)
3. Data preparation
4. Modeling (iterative, often goes back to steps 2 and 3)
5. Evaluation (from metrics back to the business objectives)
6. Deployment (deployment, monitoring, maintenance, how do stakeholders access the results)

Part 4: Challenges specific to AI products

• High uncertainty and the need for experimentation
• Probabilistic outcomes: understand your fault tolerance
• Evaluating the ML model
• Data issues: collection, annotation, storage, transfer, privacy
• Model training infrastructure & cost
• Monitoring and keeping the ML model up to date
• Bias
• Explainability

‍