India is strong market for Aerospike, plan to expand to other Asian countries: Srini V Srinivasan

Databases play a crucial role in India's evolving technological landscape, powering everything from government record-keeping to streamlining business operations. With initiatives like Digital India and the rise of UPI transactions, the country's adoption of databases has gone up manifold. Aerospike, a real-time NoSQL database company, stands at the forefront of this digital revolution.  

In an exclusive interview with Srini V Srinivasan, Chief Technology Officer (CTO) and Founder of Aerospike, we delve into how its cutting-edge technology addresses the limitations of traditional databases, particularly in managing real-time applications at scale. Srinivasan also explained how the buzz around AI has impacted Aerospike's approach and the company’s recent infusion of $109 million aimed at boosting their core transaction analytics and AI solutions, including Vector and Graph databases.  
Edited excerpts.  
 
Can you briefly overview Aerospike and its evolution since its founding? What are the company's main offerings? 

Founded in 2009, Aerospike aimed to create a database that addressed issues other databases did not, focusing initially on enhancing the read-write experience for internet and mobile use cases. One of our first customers was InMobi, an ad tech company in India. 

Over time, we've expanded our features and customer base, including financial services, telcos like Airtel, payment companies like PayTM and PhonePe, and various Indian banks. Our mission is to continuously improve our offerings and serve diverse industries. 

Aerospike is a real-time database designed for high throughput and low latency transactions, ideal for the consumer economy. For example, UPI payments in India, which handle millions or billions of transactions daily, benefit from our low-cost, high-efficiency system. 

Our Hybrid Memory Architecture uses DRAM and SSDs, alongside parallel networking and CPU-based processing, to deliver millions of transactions per second with sub-millisecond latency at a low cost. 

We support various use cases, including key-value, document, and graph applications. Our platform includes a graph database, a document database, and soon, a vector database. 

Additionally, we integrate technologies like ElasticSearch, Spark, Kafka, and Pulsar for streaming systems. We also support SQL queries through Presto and run on both data centers and public clouds like Amazon, Google, and Microsoft Azure, ensuring compatibility with any platform. 
 
What strategies have you found most effective in encouraging other enterprises to adopt real-time database technologies?  
  
Aerospike was founded to address the performance gaps in traditional databases for real-time applications at scale. We saw the need for better solutions in sectors like internet and mobile, particularly in high-demand areas like ad tech, and expected these needs to spread to other industries. 

In India, early adopters like Snapdeal and InMobi used Aerospike around 2011-2012. Their rapid growth and heavy workloads helped us attract large companies like Airtel and Reliance Jio. As financial services began focusing on consumer applications, we demonstrated Aerospike's superiority over traditional technologies. 

Our customer base grew to include companies like Flipkart and PhonePe, proving Aerospike's ability to handle high-performance problems at scale. Applications such as real-time decision-making, recommendations, dynamic pricing, risk analysis, and fraud detection require instant data access, which Aerospike excels at providing. 

We showcase Aerospike's benefits through proofs of concept, where users often find our system superior to their existing solutions. Once convinced, companies find it easy to adopt Aerospike, allowing for growth without technological limitations. 
One major advantage of Aerospike is its longevity and scalability. Customers in India and globally have used it for over ten years, experiencing business and load growth of ten, twenty, or even thirty times without needing technology upgrades. This enduring scalability is a significant benefit. 
 
How does the hype around AI affect Aerospike? How do AI-driven insights enhance data management or decision-making processes in enterprises?  

Recent AI developments are exciting for Aerospike because our technology has been widely used in traditional AI applications like fraud detection, recommendations, and risk analysis. These applications involved training data and creating models through statistical analysis and linear programming before generative AI (GenAI). 

Aerospike played a significant role in those areas. Now, with GenAI, especially using large language models (LLMs), AI usage has expanded significantly. Typically, GenAI systems ingest data over time. However, there's a challenge in using real-time data to make better decisions with GenAI, which is where Aerospike can excel. By enabling real-time, relevant decisions, Aerospike can be crucial in the GenAI development pipeline, providing real-time insights. We're collaborating with customers, many of whom have been AI pioneers even before GenAI's popularity. 
 
What are your views on the ethical implications of AI? How does Aerospike ensure its responsible usage in its conditions?  

Aerospike is a key technology that enables real-time capabilities to AI. Internally, we use AI with strict ethical guidelines. For instance, we never experiment with our customers' data. We have rules to keep data separate, so that one can’t take our data and feed it into all kinds of AI applications.  

One of our significant developments is the creation of a Retrieval-Augmented Generation (RAG) pipeline to minimize AI hallucinations. We are committed to implementing best practices to verify and improve AI outcomes. Recently, we integrated all our documentation into a GenAI pipeline, utilizing LLMs and RAG along with vector databases to showcase RAG-based applications. This process involves training the system to ensure high-quality results, which is crucial in AI applications.   
Quality is a big issue in these cases and Aerospike’s ability to efficiently store and access large volumes of vectors can significantly enhance the quality and speed of AI improvements.   

There's, of course, the ethical issue of whether one use AI or not, but we're not at that level because we have more enablers of bringing real time into AI applications. And there are a lot more components to AI where I think this applies more than, for example, the work that we at Aerospike do ourselves.  
 
How are trends such as 5G and quantum computing in telecom sector impacting data management and the role of databases in the near future?  

5G has had a significant impact on us already, because nowadays everyone has a device and they are all accessing applications, which essentially means that anything that you want to do, you can do it on your phone or any other device. It enables everyone to access applications on their devices, increasing the demand for personalized data services. This means businesses must efficiently service every customer. For example, one of our customers is changing the product description, depending on whether a person will see it based on who they are.  

So far, there’s been a huge influence in the spread of applications, which require Aerospike’s power. In terms of quantum computing, I think a lot of the quantum computing right now is happening at computing levels and less on the data. We're not yet there where we can use a lot of data to perform quantum computing, but that's a promising area. We also take part in certain high performance computing areas where we can leverage GPUs and other technologies. 
 
After the recent funding announcement, how are you planning to approach talent expansion?  

Recent investment was around $109 million growth capital investment from Sumeru Equity Partners and Alsop Louie partners. Essentially this is for investing in our core transaction analytics and AI solutions including Vector and Graph databases. There's a huge increase in demand for global analytics, as well as AI and GenAI. And there are also enormous growth areas, especially in India and Asia, in general. 
  
Since our initial funding in 2011, we have maintained a team in India which has been integral to our operations for nearly 13 years. We have teams both in Bangalore and Mountain View. We will continue to expand the teams of course, but also focus on marketing in Asian markets. India has been a strong market for us and we are now expanding to other countries in Asia. And I think that's also investing in talent on all aspects. If you look in some of our teams in the US and India, we have a lot of very talented engineers, because we're building these products together. Additionally, our global presence in sales, support, and technical areas will be expanded, along with our marketing efforts. 
 
Can you share a brief on Aerospike’s tech roadmap –  new offerings/ technologies or market expansion plans?  

We are excited to announce the upcoming release of our vector database, expected by the end of June. Following this, we'll introduce it on various cloud platforms as a service. Aerospike offers flexibility: customers can either manage the system themselves or have us run it for them. This dual approach will help us expand our vector database offering. 
Aerospike is unique because it combines a key-value database, a NoSQL database, and now, a vector database. It also supports graph features with standard graph APIs. By integrating vector search with graph and key-value queries, we aim to enhance AI applications. 

We are integrating with AI pipelines like LangChain, and services from Amazon and Google, such as Bedrock and Vertex. Our goal is to become a key player in the AI ecosystem, providing sample applications to show how enterprises can leverage their proprietary data. This allows companies to maintain a competitive edge without relying solely on general AI models like ChatGPT. 

Our immediate focus is on improving AI search capabilities. By combining vector search with key-value and graph queries, we aim to deliver higher-quality, deterministic results. Quality of results is our priority, ensuring that users get precise answers rather than approximate ones.