Forcast using SARIMAData ScienceE-commerce & Retail+5
Demand Forecasting
Jun 21, 20246 min readRead
Back
June 27, 2024
Enterprise Ops Platforms
AI as a Service
pfactorial
Share
We are living in a world where revolutionary AI models like chat GPT are transforming the whole tech industry.Every organisation is searching for ways in which they can integrate AI into their products for a better user experience,to avoid manual and tedious work and for many other reasons.Every organisation cannot afford to hire an AI team to integrate AI into their products or they might not have enough industry knowledge to understand the starting point of their AI journey.The solution for this situation is to use an AI as a service platform which offers ai models for a price, thus an experienced software developer without AI knowledge can easily integrate the ai model with the company’s product using the API provided by the AIaaS platform.
Artificial intelligence as a service provides AI tools or models as a service to an individual or an organisation.The service will be typically hosted in cloud, and users can use the service through an api or a user interface.AIaaS platforms helps organisations in building their AI products, eliminating the need to train the models from scratch.Some AIaaS services also enable the customer to train their model from scratch by providing the necessary hardware resources.
Amazon sagemaker is an AIaaS platform.Using the platform’s AutoMLfeature where we can specify the type of machine learning task including the target and features in the dataset.Sagemaker will take care of the rest including hyper parameter tuning and selection of the best algorithm for the task.
When and why?
Consider this scenario where a supermarket needs to categorise their customers into different categories based on their purchase history.The supermarket has a database of customer’s purchase history and other details.The supermarket can use this data for categorising their customer using machine learning.If the supermarket owner lacks sufficient knowledge about AI to approach a company to create a machine learning model to categorise their customers he can use a AIaaS platform like H2O driverless AI for creating a customer categorisation model.The platform will take care of finding relevant features from the data that can be used for training a customer categorisation model.
After finding the relevant features the platform also does required transformations in the dataset and trains the model using the most suitable algorithm for the task.The platform provides an option to deploy the models to the cloud using a rest api endpoint.
Types of AI as a services
● Cognitive tasks- Platforms that provide AI services related to natural language processing tasks like chatbots,computer vision,emotion detection etc ..For instance Dialogflow is an NLP product from google which can be used to integrate chatbots or voicebots in to an app
● Data labelling - platforms that provide services for labelling data.Scale AI is one of such platforms.
● Art - Text to image generation models that generate images based on text input. DALL-E from Open AI is an example for text to image generation models.
● AI model training : Some AIaaS platforms provide infrastructure for training custom AI models from scratch or fine-tune the models using customer’s data. Amazon sagemaker is an example.
● Healthcare - AI can be used in health core various tasks including early disease detection, disease classification etc. Path ai is one such platform
Creating your own AIaaS platform
Let us create our own AIaaS api using python.We will be creating a recipe generator api for a given list of cooking items using openai api.The api provides free access to openai models.For this api we are using a model called text-davinci-003,which is a large language model and flask framework for creating the api.
The get_recipe function calls the openai api and it takes cooking item names as arguments.The arguments are passed to the custom prompt and the model ‘text-davinci-003’ takes the prompt as input and returns the recipe.
Used python requests to send a post request to our api endpoint.
Results:
Benefits
- Deep knowledge of AI is not required - Organisations can integrate ai models into their products using apis provided by the AIaaS platform. For instance an image tagging app can integrate the image tagging api provided by imagga into their products instead of building an image tagging ai model from scratch.
- Reducing cost - An organisation can reduce the cost of hiring an AI team by using an AIaaS platform .
- No need for powerful GPUs - Large AI models require powerful hardware to work.With AIaaS organisations does not need the hardware to run the AI models as the model is hosted in the cloud of the AIaaS provider.
- Easy integration - The apis provided by AIaaS providers are easy to use and some providers even have libraries in different programming languages for accessing their service.
Challenges
- AIaaS provider dependency - If a company’s core product is dependent on the AIaaS provider’s service, the product’s sustainability is at risk because of the dynamic nature of AI models which may undergo changes and the one who integrates the api should be aware of this.
- Data quality - Quality data is essential for building AI models.If a AIaaS provider uses models built using poor quality data ,the API provided by AIaaS provider is of no use.
- Scalability - If a company’s product uses an AIaaS provider’s api and they want to scale their product to handle millions of requests,then the AIaaS provider’s API should be capable of handling such a volume of requests.
Future
The future of AI as a Service looks promising as more businesses are moving towards integrating AI into their products.The AIaaS providers are expected to bring new AI solutions to a variety of industries including healthcare and finance. According to a survey by software company freshbooks, 44% of small businesses expect to hire fewer people due to AI.
AIaaS platforms and their use cases
● Monkey learn
Monkey learn is an ai as a service platform that provides services including instant data visualisation ,sentiment analysis,survey analysis etc
● DataRobot
DataRobot is a platform that can be used for building custom machine learning models.It also has the capability to connect with different data platforms like google bigquery,amazon s3 etc.
● Azure AI services
Azure AI services is an AIaaS platform that provides different apis and models for machine learning tasks. IT has services including openai services(for accessing open ai models),computer vision,text translation etc.
