The Generative AI (GenAI) wave is upon us. Every enterprise is buzzing about its potential, and investments are skyrocketing. But amidst the excitement, a crucial question remains: What truly separates successful GenAI projects from the rest?
In a recent visit to India, Jihad Dannawi, Regional VP – APAC, DataStax (a leading real-time vector database company) sheds light on the hidden complexities of AI projects. This exclusive interview with CIO Dimension delves into:
- The power of the right database in unlocking the true potential of Large Language Models (LLMs)
- Why developers are the backbone of GenAI and how to empower them
- India’s unique advantage in leading the GenAI revolution
Edited excerpts:
DataStax positions Astra DB as the ‘go-to’ database for GenAI applications. Can you explain why Astra DB stands out as the best choice for GenAI applications compared to your competition?
For starters, we are a 14-year-old company that has built the most widely used enterprise-grade database. It has been delivering high scale, high performance, and high availability — even before Generative AI came in.
We are at a stage where a lot of organizations are scaling GenAI from PoC to production. At PoC level, organizations are not looking at scale or performance. But when they move to production, they suddenly start to have performance issues. They need to be able to pull in massive amounts of data to increase the relevancy and accuracy of the models, which means that the database must be bigger.
But volume often comes in the way of performance at the production stage. When organizations move beyond entry level scenarios to more mature use cases, they need to have a database that can handle their speed, scale, and performance requirements – all at a reasonable TCO. A lot of databases in the market struggle to balance these aspects.
An industry benchmark clearly demonstrates that DataStax’s Astra DB Serverless (the cloud-native database service built on Apache Cassandra) significantly outperforms the competition in throughput, latency, relevance, and TCO.
Another key offering from DataStax is RAGStack, which claims to reduce the complexity associated with Gen AI for developers? Could you elaborate?
When it comes to generative AI, one of the key issues that the developer community faces – especially when they start building the big stuff for production – is that they don’t know where to start. Because there are so many components and tools available in the market, it’s hard to decide what to use – especially when dealing with vector databases.
That’s why DataStax built a layer for the developers called RAGStack. This helps them build production-grade retrieval augmented generation (RAG) Gen AI applications in a simpler and faster way – without the developers having to learn everything from scratch. It’s fully open source, connected and improves developer productivity significantly. It drastically simplifies the process of implementing RAG – the process of providing context from outside data sources to deliver more accurate LLM query responses.
RAGStack is part of our vision is to enable and empower developers to build GenAI applications in production more efficiently. We are helping developers and companies to quickly innovate. Developers do not come with AI skills; they don’t know how to build LLMs. We firmly believe that the GenAI journey of an organization starts from the developer, and not the AI team. With RAGstack, they can start right away, without waiting to acquire new skills.
Another major concern that organizations must deal with is the AI hallucination. How is DataStax working to ensure fewer hallucinations?
Hallucination, or in other words the “context” for an LLM, is the tricky bit. Now technology allows you to launch GenAI very quickly, but GenAI is not always accurate. This is why retrieval augmented generation (RAG) is so important for enterprises building GenAI applications. They must be able to augment an LLM with relevant, up-to-date data so it can provide relevant, factual responses.
More data means better context. And this will also bring in the big competitive differentiator into the models and the eventual outcomes for businesses. This further emphasizes the point that many databases in the market that claim to be vector do not have the capacity to perform high volume data.
Can you share some examples of how Astra DB has helped customers achieve significant business outcomes?
Through Astra, we have started to discover a lot of customers with smaller needs, who can move very quickly through the GenAI journey. When customers begin to deploy impactful GenAI in their business processes, we believe that we stand to gain.
A great case in point is Skypoint, the Portland-based Gen AI healthcare company. They leverage Astra DB’s vector search to enhance their services, offering users real-time, contextual, and personalized AI agent solutions.
Another key customer for us is India’s leading edtech player, Physics Wallah. This unicorn’s app has over 15 million downloads as of January 2024. Its personalized AI bot is built on Astra DB and assists students with their academic and support queries in a personalized manner. The database’s stability helped the company handle 50x traffic increase with zero downtime during a live demo with 1 million students.
A lot of organizations are still in the experimentation stage. Our study shows that only 4% of GenAI projects have moved to production, so we are in the absolute beginning of the wave, a lot of the projects in PoC will move in to production this year.
India’s GenAI market is predicted to reach a massive $17B by 2030 and India is reportedly ahead of the US in AI adoption by businesses. What will be the catalysts for this growth and what are DataStax’s plans to capitalize on this opportunity?
AI will impact the top line and bottom line of every business out there. India has a strong community of developers and GenAI starts with developers. So, India has the skills and the mass to innovate and drive more use cases. They also have the right mindset – the ecosystem here loves to build and that’s a very positive mindset to have when it comes to AI. Because AI is not something you buy and turn on, it’s something you build. And with the right mindset, India will be faster to innovate. The GenAI wave is poised to trigger a paradigm shift, comparable to the impact of mobile and web technologies. India has the skills, capacity and the mindset to really benefit.
From a DataStax point of view, India is a big market and we are definitely investing here. We are expanding our team to meet to the growing needs.
