As we have mentioned in our ML & AI: From Problem Framing to Integration, problem framing is crucial because a well-defined problem sets the direction for the entire AI project, influencing data collection, model selection, and the overall success of the solution.

In this blog, we will expand on the question framework we used in the problem-framing phase with sample answers for our Tanoto project.

Key considerations for Problem framing Project Motivation What is the problem you want to solve?

The first crucial step in any AI/ML project is to clearly define the problem you want to solve. A clear vision of the project also gives your team motivation and values the importance of the project. Problem Definition This involves identifying the specific issue, understanding its context, and outlining the desired outcomes. This step will help you to get the right strategy goal and avoid any overcomplexity.

What specific output do you want to predict?

The clarity gained from defining the specific output is essential for making informed decisions about data collection, model selection, and evaluation metrics. Different outputs may require distinct approaches, and having a precise target allows for a more tailored and effective development strategy. What input do you have?

Focus on understanding the nature, source, and characteristics of the data that will be fed into the AI system because it may affect how your model is used for learning and making predictions. How many training samples can you provide?

If you are building your dataset, how many training samples will the dataset have? You can also add more samples by crawling the internet or checking public datasets. Performance Measurement Are there reference solutions, like a rival company’s products or research papers?

This step provides a broader perspective and context for the AI development process. It is a strategic step that encourages learning from the successes and failures of others, fostering innovation, efficiency, and the delivery of high-quality solutions. Do you have a benchmark?

A benchmark could be a well-established industry standard, a past performance metric, or even a comparable project. The key is to select a benchmark that aligns with the nuances of your particular situation, providing a clear and meaningful basis for evaluation. This way, you not only track your performance but also gain valuable insights into areas for improvement and potential refinements. What evaluation metrics are you using?

The choice of evaluation metrics varies depending on the specific problem we're tackling. We customize our metrics to align with the objectives and complexities of each task. What is the minimum level of metrics you expect?

The minimum level of metrics you expect will decide your tuning strategy and it depends on what matters to you. What would a perfect solution look like?

This question will provide direction, set standards, and guide the AI development process. Timeline Are there deadlines to be aware of?

This helps maintain focus, manage resources efficiently, and align the project with broader business objectives. When can you provide the first result / when will the customer expect the first result / final solution?

The quicker you provide the first result, the faster you can realize potential challenges, and engage the customer throughout the process to ensure alignment with expectations and requirements.

Tien

Full-stack AL/ML Engineer