AI Engineering & Deep Learning Silent Revolution in Software Development
Vyacheslav Dodatko, CEO, Co-founder, Anadea Inc
Many DL technologies that were only available to industry giants have become accessible to the smaller companies.
Neural network processing the video-feeds from security cameras in a supermarket can help merchants to place their products in the best place to increase sales. Retailers can monitor how the showcases work and what percentage of people buy the one or the other thing, depending on the showcase. It can also detect possible shoplifters in real-time.
Wearable smart devices are irreplaceable when it comes to healthcare and wellbeing. The collected data helps to diagnose health issues early, assure timely doctor visits, and even to improve people sleeping and other activities.
But the potential of wearable devices goes far beyond healthcare. Together with video-processing, they can learn to detect human behaviors and emotions. Many startups are now building new cool apps to help people with their everyday life, like:
• maintaining better posture when working at the desk, or
• making a proper swing with a golf club, or just
• make sure they do not forget to pour another cup of hot water when brewing tea.
• making a proper swing with a golf club, or just
• make sure they do not forget to pour another cup of hot water when brewing tea.
Even more advances have happened in natural language processing. Today the state-of-the-art neural networks can understand the meaning of the written text.
As a small example, recently our DL-team took the “contact us” form on our website, which had a built-in spam filter. It was working but giving false-positives from time to time, which is marking a valid inquiry as spam. And with the help of the NN model, we can now automatically analyze and sort incoming messages into several categories, and so far the DL-powered system has correctly identified every single message (no false positives).
An important trend in app development is accessibility. It becomes increasingly important to make apps easy to use for people with impairments and disabilities. And Deep Learning comes to the rescue here. Many assistive technologies are made possible by DL-enabled processing and recognition of images, speech, and videos. Neural networks assist humans and make their lives easier.
Another is information security. Everyone needs the data to be protected, the information to be safe and personal information to stay personal. And again, DL models allowed great improvement in real-time monitoring and anomaly detection. Neural networks learn to detect fraud patterns, find anomalies in data and users’ irregular behavior, analyze and prevent attacks.
Opportunities for DL are endless. I am excited to be a part of the DL revolution. To witness how neural network models are becoming more and more advanced and capable of performing many tasks only available to humans makes me quite enthusiastic.
To harness the power of modern neural networks (deep learning) it is necessary to reconsider the whole approach to software development. DL models are not as simple as a library or service ready to be plugged into an existing application with minimal effort. For the engineering team to be able to create a great product based on AI they need to have a thorough understanding of DL concepts. To gather and prepare the dataset, to train the model (or choose the pretrained one), to validate its results, and to integrate it with the application the team needs to “think in DL” from the ground up. And consider ModelOps and DataOps to traditional DevOps engineers. At Anadea we add those operational approaches to our service offering, to be prepared for the market change.
Let me say, that for 20 years on the market, we have experienced many technological innovations and we’ve helped our customers to get the most out of them. This time we see the huge wave of mindblowing possibilities is coming, to disrupt the tech world and to change the world!
