Google Cloud Next 2021

Hologic: Transforming cervical cancer screening with GCP

Hologic, a global leader in women's health at the forefront of cervical cancer screening, is in pursuit of the WHO goal of eradicating cervical cancer around the globe.  

Hologic turned to Slalom and Google Cloud to develop Genius Digital Diagnostic System, an innovative new technology that uses a deep-learning based artificial intelligence (AI) algorithm with advanced volumetric imaging technology to identify pre-cancerous lesions and cervical cancer cells in women.

The Hologic team—with Slalom and Google Cloud—designed and launched a prototype with initial results that produced around 70% sensitivity when scanning more than 20,000 images. This result demonstrates a high likelihood that this digital algorithm performance can match the accuracy of human review. This level of accuracy is an essential measure, as many places in the world do not support that level of human care. 

The effort will begin to address common customer pain points related to data integration; image labeling and annotation; reusable and reproducible AI; and the capability to run models anywhere, on premises and in the cloud. 

The Slalom team helped facilitate collaboration between organizations to promote the Google Cloud and Slalom shared vision for medical imaging AI—to transform disease detection and diagnosis and improve people’s lives by making the world’s imaging data accessible, interoperable, and useful.  

Discover Slalom’s key takeaways for Google Next 2021

Dispatches from the field at Google Next 2021

Slalom team members attended presentations that related to four different tracks at the Google Cloud 2021 conference:

• AI + machine learning
• App modernization
• Industry transformation

AI + machine learning day 01 takeaways

Build an interactive machine learning application:   Google’s Vertex AI platform and Looker extensions make it easier than ever to build business-facing, scalable, and secure data science applications. Test and training datasets are stored in BigQuery, models are trained, tuned, and deployed through Vertex AI, and with Looker, you can create accessible, beautiful, and informative data products. Simplified and integrated technology gives data scientists the capability to have more meaningful conversations with business decision makers, which accelerates enterprise adoption of AI.

Turn “dark data” into structured data:  Google’s Document AI solution transforms “dark data” into structured data that enhances enterprise and machine learning development. Digital documents (such as PDFs with digital signatures) are integral to business processes, and largely exist outside of traditional data lakes and data warehouses. Google’s Document AI solution businesses to understand, enhance, and optimize processes in new ways.

Develop custom AI apps in simple and cost-effective ways:  Google Cloud Workflows is a YAML-based declarative language, built on a serverless compute infrastructure that orchestrates any APIs and REST services with a pay-per-use pricing model. The Workflows solution allows developers to orchestrate and integrate across a variety of business processes, such as approvals, and create applications with embedded machine learning and AI.

AI + machine learning day 02 takeaways

Some AI use cases have matured across industries:  There are common business challenges that are routinely tackled with AI solutions, such as speech recognition and natural language processing in media, personalized recommendations in retail, anomaly detection in manufacturing, and medical risk prediction in healthcare. The time, energy, and cost to deploy one of these solutions is lower than it’s ever been. Given advances in hardware (such as GPUs/TPUs) and reliable, secure, and scalable cloud infrastructure—plus services like Vertex AI—the cost to deploy a pilot solution is low, with a clear path to scale if it’s successful.

Contact center AI surfaces insights into customer needs:  Contact Center AI insights create analytical value from pieces of data that are collected in every interaction with your contact center. Topic models are automatically surfaced to agents—insights that skilled analytics professionals would have needed weeks to find. Transcribing, coding, and modeling becomes an automated process that shows up on a dashboard, with common themes across conversations highlighted. Conversation analysis allows for detailed annotation of every conversation, so that users can break down the transcript into useful data. Contact Center AI Insights is easily integrated into a Looker Block, landed in a BigQuery table using Cloud Data Fusion, or even pushed back into Dialog Flow and Agent Assist so they become smarter over time.

Safely and securely translate document and text with Google’s translation engine:  In a globalized world, translation is more important than ever for reaching your customers where they sit. By working with the Google translation engine API, organizations can get started quickly, scale securely with high availability, and even customize to control domain-specific terms or phrases.  At Eli Lilly, translation activities were cost prohibitive, too complicated, and took too long to deploy. Using Google translation engine deployed on Google Cloud with a node.js front end, Eli Lilly developed an end-to-end application that allows users to get accurate translation of text in a highly regulated environment.

AI + machine learning day 03 takeaways

Diaglowflow-powered business messages creates competitive advantage:  Consumers are used to turning to Google in their moments of need, and automated business messages, powered by AI, are a natural extension of that consumer behavior. Messaging (rather than calling) has become a norm related to brand engagement. Google’s Diaglowflow, a natural language understanding tool, which powers “Bot in Box,” enables businesses to get started with AI messaging almost instantly. Organizations can even use their existing FAQ pages to get up and running within minutes to pilot the solution. Across industries, organizations are using AI-powered business messages to connect meaningfully with consumers and meet their needs.

Vertex AI allows organizations to realize the value of machine learning investments:  Almost half of machine learning models never make it into production. Using Vertex AI as a centralized platform for AI drives meaningful collaboration, clearer models, and appropriate governance, and encourages the deployment of responsible AI with Google’s built-in “AI Principles.” A unified data ecosystem and unified data platform fast tracks machine learning innovation and ensures that more machine learning pilots make it into production. It’s thrilling to see Google build solutions while keeping business decision makers and data scientists in mind—it improves the data scientist’s or machine learning engineer’s day-to-day experience, while also making it easier for business users to make more data informed decisions.

Vertex AI Workbench, a single interface for machine learning, now in public preview: This solution is a natural  evolution of data scientists working within Jupyter notebooks on Google Cloud. Vertex AI Workbench facilitates data to training at scale, which allows for exploration and analysis of datasets in BigQuery, rapid prototyping and model development with scalable compute resources, and integration, training, and model deployment workflows in one interface. Twitter and Wayfair are already using Vertex AI Workbench for data science work, and I can’t wait for the GA release.

App Moderization takeaways

Google announced a new solution portfolio of fully managed hardware and software solutions called Google Distributed Cloud. The portfolio is built on the Anthos managed platform for deploying applications across multiple computing environments, including different cloud platforms, edge computing on mobile devices, and on-premises.

Migrate for Anthos minimizes the manual effort required to move and convert existing applications from virtual machines into native containers. With Migrate for Anthos, you can easily migrate and modernize your existing workloads to containers on a secure and managed Google Kubernetes engine or Anthos clusters.

Google is partnering with telco partners to use the power of 5G to ramp up innovation by moving computer workloads closer to the user. The result could reduce latency levels and create a whole new range of mobile experiences that weren’t possible before.

Download our "Legacy Platform to a Modern Application" article here. 

Industry transformation takeaways

One of the coolest new product offerings for Manufacturing 4.0 is Google’s Supply Chain Twin, a solution that builds a digital representation of your supply chain with end-to-end visibility, alert-driven event management, analytics, and collaboration across teams.  Share data across partners without complex integrations into each organization's environment.

My second favorite solution was Google Cloud Carbon footprint, which allows users to measure, track, and report on carbon emissions associated with cloud technology. Enterprise organizations can now track sustainability practices.

Overall, the focus for the Google Cloud Platform is on simplifying and unlocking valuable insights from complex, multi-cloud ecosystems while listening to customer pain points so that they can co-invest and co-develop industry solutions that accelerate the transformation journey.