Kailash Nadh, Chief Technology Officer (CTO) of Zerodha, who has a PhD artificial intelligence And Computational Linguistics led the development of Kite, the company’s main trading platform, and is leading the company into its technological future amid increasing competition. In an exclusive interview, he talks about using large language models (LLMs, or machine learning models that can understand and generate human language text from large amounts of data) to help with technical tasks, Which saves a significant amount of time. Extensive decentralized innovation in AI technologies is taking place in the open-source world and new breakthroughs and improvements are emerging on a weekly basis. Zerodha He says some of these open-source AI tools are being used with self-hosting to make internal back office-related organizational tasks more efficient. Part:
It is widely believed that Artificial Intelligence (AI) will impact the job market, and cause serious redundancies. Also, humans have the creative power to deal with such a situation, and they have done so in the past when machines threatened their jobs. At this point in history, where is the balance?
AI technologies are multidimensional unlike other technologies. For example, a student, lawyer, researcher, writer, and a software developer can all use the same LLM tools to find straightforward solutions to a variety of problems in their respective fields. This is very different from how common tools like word processors provide problem-solving tools. This time, I think it is different, when the idea of creativity has also become a heated philosophical debate in the context of this new set of technologies.
Of course, we have to consider common sense. We cannot sacrifice good decisions in the name of automation and efficiency. For example, in case of insurance claims. Relying on AI to make high-impact decisions is not yet a good idea, and these should remain with humans who are accountable for them. There should be some guard rails for critical areas and the rules on this are a global debate.
You said earlier that generative AI is a real breakthrough, contrary to most fads in the tech sector. why did you say that? Can you mention some technologies which surprisingly came into use later?
These technologies work surprisingly well. Language, text, speech, imagery, video, and tools powered by generative AI technologies have been commoditized in no time and have become widely available for daily use. Millions of people use them directly on a daily basis. I personally have been using LLMs heavily to assist with technical tasks, and they are saving me a lot of time, which was not possible before.
Of course, there is a lot of hype surrounding these technologies, but there are important facts behind it. There are lots of fads in technology. Remember Blockchain, which was meant to revolutionize the world? Or ‘Big Data’, which became a buzzword where every organization stood to gain untold benefits from massive amounts of data? What about 5G? It aimed to revolutionize everything from mobility to ‘smart cities’ and more.
Are you one of those people who used to be an AI skeptic and then became an AI optimist? I remember an article you wrote a few years ago, referring to the “powered by AI/ML” nonsense being peddled by snake-oil salesmen.
I am not an AI-optimist or an AI-skeptic. I was and remain a strong skeptic of the “powered by AI” claim, where organizations use that phrase thoughtlessly in an attempt to differentiate themselves, while not using any AI technology at all or some rudimentary form of it. not used. With recent breakthroughs and commoditization of AI technologies, one can easily integrate AI technologies and claim to be “powered by AI”, making the phrase meaningless.
How much of a fan of automation are you? Are there elements we should not leave to automation?
I’ve been writing software and building technologies and enjoying doing it professionally and personally for a very long time. The majority of my work involves writing software and automation that simplify humans’ lives, creating user-centered technologies that provide meaningful utility and quality-of-life improvements. Any kind of important decision that affects life or society, I would not leave it completely to automation. For example, service delivery to citizens, processing of insurance claims, etc. Such important decisions must be made by humans who can be held accountable.
You have played a key role in building Kite, the company’s main trading platform. It is known for its smooth user experience. Zerodha was a pioneer in this field but competition is increasing. How do you think Zerodha can maintain its lead, given that technology in this sector has become commoditized?
Two companies may use the same framework, programming language, same database but how they package it and ultimately deliver it to customers makes a huge difference. I have seen many of our competitors impose things on customers that are not in their best interests but generate additional revenue for the company. Many of our competitors are engaged in motivating customers to do more business and generate more revenue for the company. We don’t do that. If you open our app you will not see any product or loan pushed towards you. This is our business philosophy.
We do not have external investors. Companies that have raised VC funding have to answer to their investors and it all shows in how you package your product. There is no pressure from investors on us whereas there is pressure from investors on our competitors because they are all heavily funded. That’s our advantage and I’d like to think that our lead is growing.
It is clear that open-source has a definite role to play in generative AI. How actively is Zerodha exploring these options? Zerodha has also launched a dedicated $1 million annual fund to provide financial support to open-source projects globally.
Extensive decentralized innovation in AI technologies is taking place in the open-source world. New breakthroughs and improvements are emerging on a weekly basis. At Zerodha, we are experimenting with self-hosting some of these open-source AI tools to streamline organizational tasks related to internal backoffice. It is working quite well. With our newly launched Free/Libre and Open Source Software (FLOSS) Fund, we aim to provide financial support to critical Free and Open Source Software (FOSS) projects that are vital to the ecosystem. We have created a small dedicated team internally to drive this initiative.
Can you give us an example of how Zerodha used AI tools to bring greater efficiency to the company?
Let’s transform the quality assurance process. Our team has been listening to thousands of recorded customer calls over many years. It was a manual process that was soul-crushing for the team. We created a pipeline of calls and used Whisper, an open-source model to convert voice to text. We then used a locally hosted LLM to analyze this text on certain parameters. LLMs are now used to analyze these transcripts, and we are able to identify where quality parameters are not met without resorting to random sampling. This has resulted in huge increases in efficiency.
I am learning a new language called Rust. Started working on a project in Rust and am building a fairly complex software while learning the language from scratch with the help of LLM. I’ve created a fairly well-written prototype in this language in just a few hours, which otherwise would have taken days. As a senior engineer, when I get stuck in big problems, I send the problem to the LLM, and he gives me the solution in 30 seconds, which might otherwise take 30 minutes. What is happening now is unimaginable. Even when dealing with a complex engineering problem, the LLM may suggest three approaches, and you can exercise your faculty to choose one.
What types of job roles do you see becoming redundant in India due to AI in the next few years? And why?
The most obvious candidates seem to be entry-level tasks where language comprehension and creation are involved. Programming work by junior developers, cataloging and summarizing research material by research assistants, and corporate writing and graphic design work appear to be low-hanging fruit.
Do you see engineering students now turning to AI-related courses? What would your advice be to them?
I do, and I don’t think it’s necessarily useful for a large number of students. “Big Data” courses were also popular at one time, remember? I don’t think it was a good thing to have so many students trying for engineering courses. The only advice I can give students is to search for problems they can relate to, and work on personal projects that solve them, gaining first-hand experience. Manufacturing technologies with practical experience outperform everything else and provide a significant edge.
(Note to readers: Aye Aye is a column that deals with Artificial Intelligence and its possibilities by interacting with the brightest minds in the field)
