Building a Connected Supply Chain with Lakebase

Summary

Manufacturing operations face two compounding challenges: functional data from multiple operational systems is needed for a single workflow, and process handoffs between systems rely on email and spreadsheets because no application bridges them — a gap that Databricks connectors, Lakebase, and Apps can close.
Genie Code enables vibe-coding — assembling complex supply chain applications in hours or days rather than months by combining Lakebase for the transactional data layer, Databricks Apps for the user interface, and model serving for intelligence, all described to the platform in plain language.
Lakebase provides a managed Postgres layer for the application's transactional reads and writes, with copy-on-write branches for isolated development environments, autoscaling, and point-in-time restore and snapshot management for production reliability.

Building a Connected Supply Chain with Lakebase

Manufacturing teams often need data from multiple operational systems to support a single workflow, but the handoff between systems still relies on email, spreadsheets, and manual steps. This demo, presented by Max Köhler, Lead Solutions Architect at Databricks, shows how you can mirror operational data into the lakehouse and use it as the foundation for an application that connects procurement and production.
You’ll see how a supplier portal, warehouse clerk workflow, and manager dashboard can be assembled on Databricks using building blocks such as Lakebase, Databricks apps, Genie, and model serving. The demo also walks through Lakebase as the transactional data layer, including branches, SQL work close to the data, and backup and restore options.
Watch now to learn how:
→ To mirror operational systems into the lakehouse as a starting point for application development
→ To connect procurement and production workflows in a supplier portal for booking delivery slots
→ A warehouse clerk can confirm or reject delivery requests in the application
→ A manager dashboard can show aggregated slot utilization across the plant
→ Lakebase supports transactional reads and writes as a managed Postgres layer with branches, autoscaling, and restore options
Learn how manufacturers are using industrial AI with Databricks: https://events.databricks.com/FY27Q2-WB-ManufWebinar

Chapters

FAQs

What manufacturing problem does this Databricks solution solve?

This video addresses a common manufacturing challenge: the data needed for a workflow — such as connecting supplier purchase orders to production slot scheduling — spans multiple operational systems, but there is no application to bridge them, so teams resort to emails and spreadsheets for every handoff. The solution demonstrated mirrors SAP purchase order data into the Databricks lakehouse and builds a supplier portal on top of it where suppliers can book delivery slots, warehouse clerks confirm or reject bookings, and managers see aggregated dock utilization on a dashboard. The entire application is assembled using Lakebase for the transactional layer and Databricks Apps for the user interface.

What is Lakebase and how is it used in application development?

Lakebase is a managed Postgres layer on the Databricks Data and AI platform that supports transactional reads and writes for applications that need real-time data interactions rather than analytical queries. This video demonstrates using Lakebase as the data layer for a supplier portal, handling booking creation, confirmation, and rejection in real time while the main lakehouse continues to hold the broader analytical data from SAP and other operational systems. Lakebase also supports copy-on-write branches for isolated development environments, autoscaling, point-in-time restore, and snapshot management.

How does Genie Code accelerate manufacturing application development?

Genie Code enables vibe-coding — describing the application you want in plain language and having the Databricks Data and AI platform generate the code, wiring it to the data and deploying it — so that an application connecting procurement and production workflows can be assembled in hours or days rather than taking months of engineering effort. This video shows a solutions architect using this approach to build a three-role supplier portal (supplier, warehouse clerk, and manager views) as a demonstration of how quickly operational applications can now be assembled on Databricks. The approach makes real-time connected applications accessible without requiring a large dedicated engineering team.

How do Lakebase branches work and when should I use them?

Lakebase branches are copy-on-write isolated environments where developers can make changes to the database schema or data without affecting the production instance, using the same conceptual model as Git branches for code. This video demonstrates creating a branch, making schema changes and testing SQL queries against it, and then managing the branch lifecycle independently from the main database. This capability is particularly valuable when multiple developers are working on the same application simultaneously or when testing a schema migration before applying it to production data.

Full transcript

[00:04] Hi, I'm Max and I'm a lead solutions architect working for Databricks and I support our manufacturing customers in Germany. And this presentation is about lake base and app development on Databricks. So, let's get started. One of the strong values of Databricks
[00:16] manufacturing is that you can integrate data from different parts of the value chain. So, let's get started with the value chain of products. It starts with R&D, goes over to procurement, production, logistics, sales, and finally when the
[00:29] products are in the field, we have the after market and field services. Along this value chain, we have operational systems with data inside of it. And in this setting, we see two challenges. One is
[00:42] the functional data from multiple sources are needed for specific applications. The second is and that's somehow related, is that there's a process gap when transitioning from one operational
[00:55] system to the next and we see a lot of mails, spreadsheets, and so on. So, we think we can improve on this. The chance is that the data is already there. So, Databricks offers numerous
[01:07] connectors to mirror operational systems to the lakehouse. From there, we can build on the platform by for example using apps or using Genie to have a
[01:20] conversational experience with your data or using lakehouse for the transactional data layer. But there are more building blocks on the platform.
[01:32] Nowadays, it's very easy to assemble these building blocks to a final solution because we have white coding. For example, the Genie with Genie code, we can easily white code an application within hours or days instead of months.
[01:47] So, the solution is to mirror the the operational systems to the lakehouse, then build an application on top of it for example using lakehouse, using Databricks apps, or using model serving or ends and we can do that nowadays much
[02:00] more easy by using white coding. So, let's see an example here. This specific example sits in between procurement and production when we see
[02:12] the value chain here. And the value of this specific application is that SAP purchase orders can easily be mapped to delivery slots in a production plan. So, it's in between
[02:24] these two parts of the value chain. Supplier logs into a portal, picks a purchase order, and a delivery slot for the plant and the dock, and then the warehouse clerk confirms it or rejects
[02:36] it, and there's a manager view which basically is just a dashboard which shows shows the utilization of the docks in the plant. So, what we did is we integrated a lot of data, and this time mainly SAP MM and
[02:50] some operational data in the plant. We then built an application on top of it, and we did all of that by white coding. So, we didn't code ourselves. So, let's see the application at work.
[03:04] This is the application. This is the supplier portal here. For example, the supplier number one can select my purchase order.
[03:19] Select my my my time, and then maybe I want to go to dock B at this time frame here. I can then book a delivery slot. Warehouse clerk clerk sees this request. For example, this one here.
[03:31] And then confirm it, complete it, whatsoever. And then there's some dashboard showing some aggregated statistics. For example, the slot utilization.
[03:43] Now, we used different building blocks to build this application. And in this very short presentation, I just want to um, go into deeper into one of these building blocks, which is Lakehouse. So, Lakehouse is the, um,
[03:57] transactional data layer, uh, within Databricks. Um, it is, uh, managed Postgres, basically our managed Postgres Postgres offering, and you need it when you when you want real-time transactional reads
[04:09] and writes as a basis for a responsive applications. We think that there are three main advantages. Uh, one is is it is developer-friendly. For example, using branches, um, you
[04:21] only pay for what you use. It's there's auto-scaling and it even scales to zero, and it's integrated into our platform. So, let's have a look at it. Um, I just clicked here on the persona, uh, Lakehouse Postgres.
[04:33] Postgres, um, and, um, then I see the different Postgres projects. Um, let's click on this one here. I see a dashboard here, and I immediately see,
[04:45] um, different branches. Um, so, this is the dev branch, a production branch, and a branch is, um, um, just facilitates development. So, it is an isolated pro, uh, uh, Postgres environment, uh, which is, uh,
[04:58] copy-on-write, and, um, therefore, it doesn't duplicate any, um, any data. So, it facilitates development. We can also see that there's a SQL editor. As developers, we want to work
[05:10] close to the data, um, and for example, when here, I want to click this schema here, delivery app, run the code, and I see the final table, um, which is the, um,
[05:22] confirmed slots, and, uh, I can also analyze the query, and I here I see in, uh, the, um, analysis results. I can also see the tables.
[05:37] Here I have four tables. I can filter it, edit it, and so on, to work closer to the data. Of course, there are also backups and restores, um, for this transactional data layer, so I can restore to a point in time restore and also do create a manual
[05:52] snapshots or schedule snapshots here and restore from a from a snapshot. But this is is a very short summary of Lake Base. It's just one of the building blocks we use for the application.
[06:04] Talk to us if you want to learn more about application development and about Lake Base and this content here. Thanks a lot.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.