The consistent throughline of any business looking to stay competitive in an evolving market is extracting value from data—from key trends to relevant insights, prescriptive actions, and more. This is done through data applications, acting as enablers of productivity, efficiency, and innovation across the enterprise. However, due to the dichotomy of legacy systems, data silos, and poor data quality and the demands for real-time, scalable data, building effective data apps is a challenging—yet rewarding—feat.
Experts joined DBTA’s webinar, Building Modern Data Apps: Choosing the Right Foundation and Tools, to identify the ways in which modern data apps can achieve the data goals that nearly every enterprise seeks while addressing their unique architectural inhibitors.
AI-driven, adaptive apps are the future, noted Mark Gamble, director product and solutions marketing at Couchbase. Apps that are hyper personal, contextual, innovative, responsive, and available everywhere—both online and off—are some of the qualities that enterprises and consumers alike are hungry for.
Gamble explained that, to enable these sorts of applications, Couchbase acts as the operational data layer for these apps with its cloud-native, NoSQL database. With the ability to power massive applications—such as LinkedIn, Tesco, Comcast, United Airlines, and more—Couchbase delivers a unique, multi-purpose approach that offers a combination of database capabilities, including:
- Key-value access
- SQL query with AI assistant
- Full-text/vector search
- Real-time columnar analytics
- Time series
- Event and streaming
- ACID transactions
- Mobile database, and more
“These are all data access patterns that our customers no longer have to bolt together from a bunch of bespoke, different technologies,” said Gamble. He further added that Couchbase Capella is AI-ready; meaning, through both its columnar service and vector search, Couchbase supports AI for on-device data processing as well as cloud data processing and synchronization.
Artyom Keydunov, CEO and co-founder of Cube, remarked that a Universal Semantic Layer is crucial in addressing the ways the data stack has evolved. Once monolithic, data stacks now consist of thousands of parts with semantic layers in each tool—driving inefficiencies, inconsistencies, and inaccuracies. This leads to “model chaos,” where semantic models are duplicated across different parts of the stack to maintain some modicum of consistency, explained Keydunov.
Cube’s innovation brings a model once, deliver anywhere approach where a Universal Semantic Layer sits on top of the data warehouse to persist a unified semantic principle across BI software, customer-facing analytics, and spreadsheets. Cloud-native, open source, fully hosted, and multi-tenant, Cube’s Universal Semantic Layer combines the powers of data modeling, access control, caching, and APIs to maintain a consistent, accurate semantic layer.
From the source data to the database, queries and the semantic layer, predictive models, and ultimately, the application itself, Steven Hillion, SVP, data and AI at Astronomer, emphasized that there must be coordination among each of these components.
“Businesses these days depend critically on data—whether that’s for producing applications or predictive models or dashboards, key metrics, reports, and so on—...but pipelines are hard to run in production in a way that’s completely reliable,” noted Hillion.
Part of this difficulty, continued Hillion, is due to the historical separation between the three key players of app development: the data engineer, the data scientist, and the app developer. Each of these teams work in siloed environments, hindering the potential of the app they work to create.
Astronomer solves these particular pain points with Astro—Astronomer’s modern orchestration platform for Apache Airflow that unifies data, AI, and applications. Astro ensures that data is delivered on time with the speed and scale that modern apps demand, supporting data-powered applications, critical operational processes, analytics and reporting, and machine learning (ML) Ops. With a production-first, orchestration-first mindset, Astro invites a collaborative atmosphere where data pipelines are written in a way that can be easily transferred to production.
For the full, in-depth discussion regarding the tools, techniques, and strategies for building modern apps, you can view an archived version of the webinar here.