Lecture Notes on Data Science and AI Solutions

Introduction

• Speaker: Senior Manager, Data Science at SoftBank Energy
• Focus: Renewable energy platform for developing and maintaining energy projects.
• Key Areas of Work:
  • Forecasting problems
  • Anomaly detection
  • Optimization
  • Predictive maintenance
• Partnership with Abacus:
  • Deployed models over the last six months
  • Expansion of partnership with Abacus.

Customer Introduction

• Speaker: Anthony from Johnson Controls
• Role: Runs a global team of data scientists.
• Key Projects:
  • Churn models (12 models in production)
  • Anomaly detection in sales purchases
  • Optimization project for CFO regarding cash collection.

In-House vs. Vendor Solutions

• Importance of deciding whether to build AI solutions in-house or purchase from a vendor.
• General consensus:
  • Building in-house is resource-intensive and often not feasible.
  • Preference for traditional providers (AWS, Azure) or specialized platforms like Abacus.
  • Key considerations:
    • Collaborative development effort
    • Time efficiency and support from the vendor
    • Ease of deployment in production environments.

Platform Experience

• Positive experiences transitioning from Azure to Abacus:
  • Collaboration and support are key factors.
  • Significant time savings in model training and deployment.

Integration and Infrastructure Challenges

• Integrating Abacus with existing infrastructure:
  • Most data still in Azure, integration relatively straightforward.
  • Some security protocols create barriers for data transfer.

Code-First Data Scientists

• Discussion on resistance from data scientists who prefer writing code versus using platforms like Abacus:
  • Depends on use case; some advanced or innovative projects may require custom solutions.
  • For common use cases (forecasting, churn detection), using existing solutions is more efficient.
  • Emphasis on enhancing data quality rather than reinventing models.

Generative AI and LLMs

• Perspectives on leveraging generative AI in business:
  • Improve productivity through information sharing and workflow automation.
  • Potential integration with tools like Microsoft Teams for data access.
  • Use of AI as an augment to existing processes rather than a replacement.

Favorite Features of Abacus Platform

• Efficiency in training multiple algorithms simultaneously.
• Feature importance analysis for better business insights.
• Optimization solutions that are not readily available in other platforms.

Suggestions for Future Features

• Enhancement of feature engineering tools within the platform.
• Improved decision interpretability for forecasting and optimization outputs.
• GIS-based use cases for solar energy development.

Advice for Executives on Leveraging AI

• Define clear, quantifiable outcomes before starting AI projects.
• Focus on solving specific, measurable problems.
• Adopt a crawl-walk-run approach to gradually build AI capabilities.
• Emphasize collaboration and iterative learning in the process.