At the event, Amazon announced several new capabilities and product launches across its expansive product line, focusing on AI capabilities that it plans to build on.
Some critical AI applications and services related announcements are covered below:
Explainability and human-in-the-loop workflows
The new Amazon Augmented AI service (A2I) was presented at Re:Invent. The A2I allows improved inference by allowing human reviewers to validate machine-learning predictions actively. AWS hopes to create more explainable models through human insights derived from A2I integrated with SageMaker.
Custom labels in Rekognition
Rekognition, AWS’s computer vision service for video and images, will allow custom labels now. This means that businesses will now be allowed to upload their images for domain-specific use cases.
Amazon Kendra for Enterprise Search
AWS debuted Kendra, which is a new ML-powered enterprise search service. Kendra can aggregate documents and knowledge artifacts from across siloed repositories within the enterprise. This will enable firms to create an intent-based search index that provides more contextualized, useful, and relevant answers to natural language-based queries.
AWS launched the Transcribe Medical, which is a service that applies natural language specifically to medical documentation, reducing the documentation workloads for doctors.
Amazon Fraud Detector
AWS filled an evident product gap by announcing a service that allows the automatic training and building of custom fraud detection models. Non-data scientists or resources can use this detector regardless of their less subject knowledge.
The most applauded moment at Re:Invent this year was the CodeGuru announcement. This service aimed to automate code reviews using ML. CodeGuru includes profiler and reviewer tools to catch code issues, identify expensive lines of code, and also suggest ways to streamline the coding system.
The cloud contact center suite Connect, AWS’s only business application, is designed to add advanced ML-based analytics with Contact Lens. Amazon Connect allows transcribing and analyzing customer calls for various issues such as compliance lapses, negative sentiment, and long pauses. AWS said that the company is also working towards releasing real-time call transcription services in 2020.
Compared to last year, AWS emphasized on fully embracing modern architecture this year. AWS has focused on a more pragmatic vision of digital transformation. From the perspective of product development, this year, AWS has promised more agility and convenience for enterprise developers by refining its service features.