Select Page

What is TensorFlow? All You Need to Know about the Machine Learning Library

By now you have probably heard about data science techniques like machine learning, deep learning, neural-networks and AI. You are likely aware that they are the future, but chances are if you don’t dig too deep into the field, you would not have about TensorFlow. So, what is the big deal about it, and why should you care? 

To answer the first question, TensorFlow is like the backbone of AI (well, one of the few most advanced and widely used backbones at least) as well as the exciting applications that you hear about. It is the platform on which you build, train, and deploy all those powerful AI models. Most importantly, it is everywhere: the Google mail sorting system uses it, Spotify uses it to create personalised playlist, the list goes on and on. You are almost certainly using something built on TensorFlow right now, you just don’t realise it. 

For the “why should you care” answer: if you want to explore, build, and integrate any machine learning application in your business, you should do it with TensorFlow to get the “State of the Art” results and best deployment. Don’t take my word for it, just see the choices made by Coca-Cola, Airbnb, Qualcomm etc.

User cases:

What we could do with it? What are the benefits? 

Case 1: Intelligent email sorting for customer service in Swisscom 

Swisscom developed and implemented the AI email sorting system that “records the content and sends the e-mails directly to the right recipient, without any need for manual e-mail assignment… It anonymises confidential data, recognises the language and identifies certain keywords, hot topics or VIP customers.” 

Aside from all the time and cost that an automatic system like that could save, the system also reached the success rate of sorting of 90%, which is higher than human sorting. Customer satisfaction is greatly improved by the fast and accurate responses enabled by the system.  

Case 2: Fraud detection at PayPal 

Fraud detection has always been a big issue for PayPal.  It must accurately identify frauds while not disturbing the normal customers. And with the amount of transactions and the complexity of it, it is impossible for PayPal to manually verify every transaction or have a simple decision-based system to do that. 

Introducing AI. Using TensorFlow and other frameworks, PayPal “has been continually achieving 10-20% more accuracy over other traditional ML approaches in real time fraud detection.” 

Existing integrations – How hard or easy is it to integrate this into existing businesses? What integrations are available? 

The most powerful things about TensorFlow are not just its AI model development capabilities or the large pool of APIs and existing models available, but also the capabilities that can be deployed easily across different platforms. This means that you don’t need a Python environment setup to make it work. 

Do you want to integrate AI to your company’s website?

There is a TensorFlow.js JavaScript library available.

Are you looking to put AI models into your mobile devices? 

TensorFlow Lite could squeeze the AI models into Android, IOS or other IoT devices.

Want to scale the models to a large production environment with pipelines set up? 

TensorFlow Extended provides an easy way to set it all up. From big to small and front to back, TensorFlow makes it easy for the business to actually use the AI power anywhere. 

Where to next?

Clearly TensorFlow is a powerful tool and better yet, it is also open source, meaning that it is completely free.

So why don’t more people use it?  

Aside from the fact that most people are not aware of the significance of TensorFlow, getting it to work requires very specific knowledge for understanding, building, training and testing the models and using the APIs. You need access to experts. 

At Decision Inc, we have those experts and not just in Australia but globally. The experience of our consultants, especially those in the data science practice, can provide a lot of value to our clients. Whether it is helping them with basic data challenges or more advanced analytics, machine learning and AI tasks that businesses can use today.

So, whether it is NLP (natural language processing), object detection, time-series analysis and more, get in touch with us so we can show you those next steps on your data and technology journey.

Michael Haowen Liang
Management Consultant
Decision Inc. Australia

Latest Blog

Integrated AI Augmented Enterprise

Integrated AI Augmented Enterprise

BLOG | board | ARTIFICIAL INTELLIGENCE Integrated AI Augmented EnterpriseEnhance critical decision making and actions to deliver predicted business outcomes with Artificial Intelligence. Become an Integrated AI Augmented EnterpriseUndertaking some form of...

Get In Touch

Contact us to find out more about building apps with Power BI.