A naukri.com initiative
Medium
2w
100
Image Credit: Medium
AI playbook: My lessons learned from AI development
- In the development of AI solutions we begin with scoping, this is where we set out to understand the user’s problem and whether AI is the appropriate solution.
- If AI is used as a solution, it should be used where it can provide the most impact such as through improving the customer experience or employee satisfaction.
- When developing a solution, providing additional guidance as potential blind spots occur could lead to higher performing solutions.
- In testing, breaking out data into different categories ensures that the less frequent scenarios are adequately represented.
- When troubleshooting AI solutions, broader troubleshooting concepts can be useful in breaking down complex processes into smaller steps.
- During iteration, the initially scoped requirements can be traded for a less sophisticated and more effective model.
- Monitoring the performance of AI models post-deployment is critical to ensuring the expected results are still being produced.
- Using feedback mechanisms such as a 5-star rating scale allows us to find out early when shifts happen and not be caught unaware.
- The quality of results AI produces depends on the people involved, as development is part art, part science.
- Sharing experiences and discussing best practices can help accelerate progress with AI technology.
Read Full Article
6 Likes
For uninterrupted reading, download the app