RE:INVENT 2020 SESSIONS PICKS FOR DEVELOPERS AND MACHINE LEARNING PRACTITIONERS
This year will be remembered for many reasons. It has been a year of big changes in our lives and habits and a time when we found new ways to do the things we love. 2020 will be remembered as the first year without Amazonians from everywhere gathering in Vegas for the traditional re Invent.
Luckily, it won’t be a year without re:Invent because AWS decided to shift the conference completely online, with a catalog of almost 2000 unique sessions ranging from IoT to machine learning applications to infrastructure and serverless.
As a developer and a machine learning practitioner, digging into such a vast list to extract the best sessions to watch while doing our everyday job is not easy. Here this article comes in handy, trying to enucleate the talks you can’t miss, grouped into four main themes, with a bit of context to support deep diving into each topic.
Digital Customer Experience
Delivering tailored experiences to customers at the right time is the holy grail of marketing, retail, and almost every brand out there. Personalization requires acting on two sides: understanding and engaging our customers.
Customer understanding requires analyzing a huge amount of data. Most of them without a clear categorization: e-mails, surveys open answers, phone calls to contact center are just a few examples. If getting deep knowledge of customers from unstructured data is your go-to, you can’t miss "Discovering insights from customer surveys at McDonald’s" and "Using AI to automatically find insights in an email: A rapid prototyping story," respectively by McDonald’s and Liberty Mutual explaining how they used Amazon Comprehend to extract meaningful topics from unstructured customer feedback. The latter is also presenting a customer journey workflow, using Amazon Lex to interact with this amount of data. The best part of the story? You don’t have to be a data scientist: every developer can start implementing such features because they are available in managed AI services! Such services could also be tied together to build a machine learning-powered customer analytics solution starting from Amazon Comprehend recognized entities. Going one step further, a complete managed solution for contact center with AI to extract meaningful data has been released last year.
The other leg of Digital Customer Experience is tightly coupled to engagement. It can be achieved using managed AI services such as Amazon Polly and Amazon Rekognition as discussed in "Recreating Colonel Sanders for KFC Canada using Amazon Polly Brand Voice" with KFC making customers able to listen to the historical voice of the brand and in "Role models: How AI is improving diversity in fashion" to feel people comfortable with a more inclusive approach by Lalaland. I expect to see great real-world use cases in these talks.
On the experience tailoring to a brand-loyal customer, definitely the session "Deliver viewing experiences for super fans with Amazon Personalize" can light some interesting applications of another service ready to be used in production with no prior data science expertise.
Insights, applications, and use cases to inspire developers and ML practitioners
Following the practical applications track, we could outline two interesting topics that shine out re:Invent 2020 sessions: financial services and health.
On the side of the financial services, a concrete use case of managed fraud detector service is Amazon itself, as presented in "Catch more potential online fraud faster with Amazon Fraud Detector." At the same time, insurtech could benefit from document recognition and OCR using Amazon Textract, with human support were required to improve accuracy, thanks to Amazon Augmented AI, as discussed in "Intelligent document processing for the insurance industry."
A deep focus has been put on the latest advancements supporting medical divisions and personal health during this year pandemic, which pushed the boundaries of the state of the art forward either in the industry "COVID-19: Identity verification & work safety with Amazon Rekognition" and with patient management, with the whole set of AI services applied to clinical data management and presented in the talk "Using AI to automate clinical workflows" such as Amazon Transcribe Medical, Amazon Comprehend Medical, and Amazon Kendra in addition to the already mentioned Amazon Textract and Comprehend.
Operations support for developers bridging machine learning models in production
In recent years, developers focused not only on managed AI services but also on bringing Machine Learning models into production, thus bridging the gap between data science and software engineering. Building a machine learning project is a complex task that requires a lot of expertise in model-related stuff and in every aspect of operational tasks that support data scientists, the so-called MLOps.
AWS has tried since day one to empower companies with tools such as Amazon SageMaker. A great starting point is outlined in the session "From POC to production: Strategies for achieving machine learning at scale," "How to use fully managed Jupyter notebooks in Amazon SageMaker," and "Get started with Amazon SageMaker in minutes."
A focused talk is also "Implementing MLOps practices with Amazon SageMaker," which offers a broad overview of the Amazon SageMaker platform and its capabilities supporting every MLOps workflow phase.
What about advanced sessions for ML experts?
Data scientists looking for deep diving model training techniques won’t be disappointed by "Build quality ML models easily & quickly with Amazon SageMaker Autopilot," "Train and tune ML models to the highest accuracy using Amazon SageMaker," and "Interpretability and explainability in machine learning" while developers looking for something more tailored than managed services. Still, without extensive ML knowledge will be glad to understand the reasons behind "Choose the right machine learning algorithm in Amazon SageMaker" and how they can choose between 17 different models to be used in their applications.
Conclusions
This year’s conference is by no means smaller than previous editions. Instead, AWS took the opportunity to involve many local customers and partners to hold speeches in their native languages, such as Spanish, Japanese, and Italian. It will be a great conference for all, continuously raising the bar in the cloud and machine learning domain.
Have fun. Build great things, make history!
