I want to share a high-level overview of what we’re currently working working on at MyEvie. Training an AI model is a complex process that involves several steps. Here's a simplified overview of what we’re doing .
We want our model to be able to converse, in a human-like way, with our users, enabling them to feel comfortable talking about any issues they may be facing.
We are continually gathering data from many sources and feeding these into our AI model which we hope will lead to Evie being able to help our users
We’re busy cleaning our data, which is very time-consuming, and transforming it - extracting the relevant chunks of data.
We then divide our dataset into training, validation, and test sets. We’re going for a split of 70% training, 15% validation, and 15% test.
5. We then train the Model:
We then use the training data to train the model. The model will make predictions and compare them against the actual labels, adjusting its parameters to minimise the error.
We’ll then use the validation set to tune hyperparameters and prevent overfitting. Overfitting occurs when the model performs well on the training data but poorly on unseen data.
After tuning, we’ll evaluate the model's performance on the test set to estimate how it will perform on real-world data.
Based on the performance, we’ll go back and gather more data, try different preprocessing techniques, and tune the model further.
The good bit! Once we’re satisfied with the model's performance, we can deploy it to a production environment where it can start making predictions on new data.
We need to then Continuously monitor the model's performance over time, as data and real-world conditions change. Periodic retraining might be necessary.
It's important to note that this is a high-level overview and that each of these steps is significantly complex. Additionally, we always ensure data privacy and prevent bias in our models, which in itself is a huge task.
I hope this gives you a flavour of what we’re doing and helps you understand what’s going on behind the scenes.
Mike Georgeson
Co-Founder
