CrazyCloud - Fotolia

How MongoDB is driving AI adoption in ASEAN

MongoDB field CTO Boris Bialek delves into how the company is supporting ASEAN organisations in their AI journey, highlighting the importance of data governance and real-time capabilities in driving successful adoption

With data being critical to the success of artificial intelligence (AI) initiatives, database suppliers such as MongoDB have been doubling down on efforts to help organisations unlock the full potential of AI.

In an interview with Computer Weekly, Boris Bialek, field chief technology officer for industry solutions at MongoDB, sheds light on those efforts, highlighting the database’s flexible document model and real-time data streaming capabilities while addressing data sovereignty requirements.

He also discussed MongoDB’s AI-focused initiatives such as the MongoDB AI applications programme (MAAP) that provides practical guidance and resources for organisations ready to take the next step in their AI journey, the challenges posed by legacy systems and the importance of developer-friendly tools that cater to local languages and empower businesses to build applications tailored for the ASEAN market.

Editor’s noteThis interview was edited for clarity and brevity.

Talk to us about the work MongoDB is doing to support ASEAN organisations in their AI journey

It’s important to understand that we’re still in the early stages of AI adoption. Many clients are experimenting, but nonetheless we see the first production implementations happening as well. For us, and this may sound self-serving, data is core to achieving good AI results. It’s a classical thing from the analytics days – if you have garbage data in, you get garbage analytics out. AI is not any different – if you don’t get good data integrated, you get hallucinations and wrong information.

What we have at MongoDB is the document model and we added Vector Search capabilities last year. This allows us to accumulate in the same data set, operational data, metadata, Vector Search data and additional data generated from interactions, which we can put into context with an LLM [large language model] and connect to AI applications in real time. This is our claim to fame and how we differentiate ourselves.

We have a large European retail customer that is storing all their product information in 23 different languages from 15 data sources in one big database, available in real-time. People can start asking questions about calorie counts, for example, and it’s a great solution for consumers who would otherwise have to look for the information on the internet. We also do the same for banks in client relationship management.

We’re seeing more organisations in ASEAN with the mobile-first mentality and aggressively leveraging our capabilities. They’re more willing to try things out, while those in Europe are probably the most conservative, and North America is somewhere in the middle
Boris Bialek, MongoDB

We spend a lot of time using AI internally as well. We have millions of databases in operation on our cloud platform, and we’re using AI to optimise them for our clients.

Another area is modernisation – 30% of legacy infrastructure accounts for 80-90% of the cost. During our last earnings call, our CEO announced that a global insurance company is working on moving to our platform in 56 countries.

I’d like to dive into the recent generative AI (GenAI) announcements from MongoDB, particularly the MAAP which aims to help companies build GenAI applications faster and easing some of their challenges. But many companies are still experimenting with GenAI and haven’t scaled up their deployments. What’s your take on that?

You’re right on point. MAAP is there to help them move from experimentation to production. There’s always this question of how to start with dozens of vendors knocking on their doors with different offerings, whether it’s LangChain or Llama, so clients get a little confused. We deliver a more curated approach with certified and qualified reference architectures. We help them with data integration and integration with various components like an embedder and a LLM gateway.

We are also working on a book to define use cases to help clients jumpstart their first projects. There’s a lot of interest in ASEAN around that because many companies, as you pointed out, are experimenting. They’ve come to the point when they’ve experimented for a year and are asking about where to go from there. That’s where MAAP comes in.

What about streaming data?

We’ve just announced our Atlas streaming product which allows you to stream data into the database, process the data in real-time and take immediate action. For example, let’s say you’re paying for something at a store and the cashier system notes that you’re a special client and offers you a two-for-one deal. That would be a real-time AI interaction.

But to do that, the system needs to know you’re at the cashier, what you’re buying and your interests. This is where MongoDB’s real-time capabilities come in. We can embed data, vectorise the information and build real-time vectors. We can even have multiple vectors inside the same document about you as a consumer, your purchases and your loyalty status, and ask an LLM for real-time inputs based on those vectors.

For that to work, real-time data streaming is key. A lot of AI people came from the analytics and enterprise data warehouse world, and this is completely different now that you can have a database that helps you make a decision in 50 milliseconds.

How is MongoDB addressing the need for AI governance?

What we are mostly seeing right now is data governance, specifically data lineage that tells you where a vector came from. For example, for a bank doing credit scoring to say they have 26 databases, 12 copies of the databases and data from an ETL [extract, transform, load] process sends a bad message to regulators.

With MongoDB, they can bring the data together in one document. Inside the document, we can use embedders and build a vector out of that. So, if somebody asks how the vector was generated, we can say this is the inference model XYZ that uses the following data sources. This removes a lot of ambiguity, especially for those in the banking and insurance space.

How is MongoDB addressing the needs of customers with data sovereignty and data residency requirements?

We are probably the database with the most coverage around the globe. We’ve also announced that we’ll bring MongoDB Atlas Search and Atlas Vector Search on-premise. When you have data lineage, the next requirement is data residency and sovereignty.

In Europe, we’re already working with regional sovereign cloud providers like OVHcloud, StackIT and Bundescloud, the German government cloud. We’ve worked in multiple countries with these kinds of systems, and we are very proud of it. We can also do on-premises deployments for clients where they get a self-managed and controlled environment.

One of the challenges in the APAC region is the delayed availability of new services from global technology suppliers. Does MongoDB have a release cadence for new capabilities?

Everything goes out immediately. When we rolled out Vector Search, we had it immediately available globally. The health ministry in Indonesia were very happy to see that they had Vector Search capabilities on MongoDB Atlas. With in-country support for the Atlas platform, it’s easy for customers to add and switch on a feature like Vector Search without redoing the security qualification.

There's always this question of how to start with dozens of vendors knocking on their doors with different offerings, whether it’s LangChain or Llama, so clients get a little confused. We deliver a more curated approach with certified and qualified reference architecture
Boris Bialek, MongoDB

It’s a different discussion if you use a third-party solution or you have LLM integrations outside your country. This is also why we see clients moving to local LLMs, with people training their own models in local languages. And MongoDB works with all of these components out of the box. We don’t need to build anything special because of the document model.

Most people don’t realise that most of the work we do is always multilingual under the hood. We have a developer tool called Compass where you can use natural language to ask questions or perform tasks like building a function to show the top 10 users of a product. You can do this in Bahasa Indonesia, Mandarin and English, and it gives you code afterwards. You can then add the code into your code base. This is where MongoDB is very open to developers.

What challenges do you see customers facing in Asia that might put them at a disadvantage compared to those in other parts of the world in terms of leveraging MongoDB’s capabilities?

We’re seeing more organisations in ASEAN with the mobile-first mentality and aggressively leveraging our capabilities. They’re more willing to try things out, while those in Europe are probably the most conservative, and North America is somewhere in the middle. When I see what some of our clients are building in ASEAN, to be honest, they’re not behind. Take AirAsia, for example. The environment for their consumer app and consumer interactions was built with MongoDB technology, and they have one of the leading apps, in my opinion, globally.

I think Europe is the most behind right now, although it’s ahead with regulations, so we should give them credit on that as well. With the new Dora law, there are very rigid rules around usage of the data of individuals and so on.

But on the application side, what we see in ASEAN is very positive. People see IT as a means of doing business and that helps us as well. There’s a lot going on and people are very focused on moving out of legacy systems to enhance customer experience because the market is very competitive.

Read more about database software in APAC

  • Neo4j is targeting GenAI workloads in the fast-growing APAC market by leveraging knowledge graphs to improve the accuracy and explainability of large language models.
  • Snowflake’s regional leader Sanjay Deshmukh outlines how the company is helping customers to tackle the security, skills and cost challenges of AI implementations.
  • Australia’s IAG is using Kafka’s data streaming capabilities to integrate disparate data sources and provide real-time data services to support its business.
  • The benefits of Hana Cloud Services are clear to APAC customers, although actual adoption will boil down to cost.

Read more on Database software

CIO
Security
Networking
Data Center
Data Management
Close