Brief Summary
This session by CFA Society India, featuring Tarun Dua from E2E Networks, discusses AI adoption in the BFSI sector. It covers the shift from CPU-centric to GPU-centric computing, AI use cases in BFSI (like risk management and customer experience), the importance of data privacy, and advice for young entrepreneurs. Key takeaways include the transformative impact of AI, the need for robust AI infrastructure, and the importance of data security.
- AI is transforming BFSI sector by improving customer experience, risk management and automating tasks.
- GPU computing is more cost-effective for AI than CPU.
- Data privacy is paramount, advocating for self-hosted AI solutions.
Opening & Introduction
The session is an introduction to enabling seamless AI adoption in the BFSI sector, featuring Tarun Dua, the founder and MD of E2E Networks, a pioneer in AI compute infrastructure in India and the first cloud partner of Nvidia in India. Tarun's background as a computer science engineer and his leadership in building scalable AI/ML solutions using open-source technologies are highlighted. The session aims to explore how E2E Networks provides self-service public cloud infrastructure and platform tools to facilitate AI transformation for businesses.
Brief comparison: old tech world vs. modern tech ecosystem
The old IT world was designed for CPUs with limitations that led to software minimizing data sent for processing. CPUs, while powerful, are underutilized in modern AI/ML compared to GPUs. GPUs offer significantly more cores and high-speed memory, making them cheaper and more efficient for running AI tasks. The shift to GPUs enables the use of the 99% of organizational data previously unutilized due to CPU limitations, processing visual and language data for better insights.
Context-setting note on “AI in BFSI”
The discussion shifts to the relevance of AI in the BFSI sector, emphasizing the importance of understanding business details. AI can help make sense of the vast amount of detail that human brains can't fully process due to constant information bombardment. Reasoning systems are crucial for BFSI, and AI can assist in determining the significance of information from various sources like news cycles and visuals.
India’s Cloud and AI Readiness for Financial Innovation
AI systems can provide consistent user experiences across all customers, unlike traditional systems where top customers get preferential treatment. AI can summarize lengthy documents, such as insurance claims, and remove human biases in processing them. For instance, AI can detect inconsistencies in insurance claims, like mismatched car colors or engineered accidents, and improve the accuracy and speed of processing home loan applications by reviewing documents and identifying weak signals.
AI Use Cases in BFSI: Risk, HR, underwriting, personalization
AI can determine loan eligibility, credit amounts, and repayment likelihood. It can also automate answering customer questions via email, improving response times and accuracy. Sentiment analysis helps brands understand customer satisfaction levels throughout their journey, enabling customized sales activities. The BFSI sector is particularly concerned about IT security, including data protection and threat detection.
AI Agents and Infrastructure Scaling
AI can enhance learning programs within organizations, ensuring consistent brand experiences across different locations. It can also handle transactional customer service inquiries in the first 60 seconds, freeing up human agents for more complex issues. Training specialized AI experts can improve back-end activities and overall efficiency. The key difference between rule-driven and AI-driven software is the ability to swap out models easily with the right AI infrastructure, allowing for rapid anomaly detection and resolution.
E2E's TIR AI/ML Platform
E2E Networks has built the TIR AI/ML platform to quickly facilitate the AI journey for organizations. This platform marries the work of data scientists with real-world feedback from application deployment, allowing for continuous training and improvement of AI models. The platform aims to build B2B enterprise systems, focusing on practical applications rather than theoretical problems.
Impact of models like DeepSeek on AI adoption
DeepSeek positively impacts the industry by reducing the cost of AI, enabling organizations to consider more use cases within the same budget. It advances open-source technology, delivering the promise of AI to more people. However, DeepSeek is not easily replicable outside specialized teams and requires expertise in low-level coding. Despite misconceptions, it represents a massive advance in reasoning models and AI science, with rapid technological changes expected in the open-source world.
Privacy concerns with AI implementation
Privacy concerns are similar regardless of whether AI originates from China, the US, or India. Using proprietary, non-self-hosted AI systems can lead to a loss of competitive advantage. There's a need for singletoned SAS running on your own compute infrastructure to keep data within organizational boundaries. Personal domain AI is essential to prevent leaking company knowledge to external systems like ChatGPT, requiring a close look at IT supply chains to ensure data security.
AI Agents: Power, adoption, and use cases
AI agents will lead to the disappearance of certain jobs, but they will be replaced with more creative roles, increasing overall productivity. While this change may cause temporary pain, avoiding it will lead to greater pain in the future. The world will not stop for countries that resist AI implementation, so accepting short-term pain is necessary for long-term benefit.
Government’s push and initiatives for AI
The starting point for advancing AI adoption is education. India's elite spend billions educating their kids abroad, and AI has the potential to change that. Reskilling people, including those already in jobs and kids who are going to learn, is crucial. The biggest imperative for any government is to ensure that the country does not get left behind in AI.
Top 3 AI use cases in wealth and investment management
For wealth managers, the top three AI use cases are ingesting and making sense of large amounts of data, searching for information efficiently, and personalization. AI can provide quantitative and qualitative information with reasoning, helping customers understand the implications of new developments. For example, AI can assess whether DeepSeek is good for AI infrastructure and application development, providing validated reasoning.
Advice for young professionals on entrepreneurship
Entrepreneurs see things that others do not, and the cycle time for doing anything is decreasing with new technology. Despite India's lack of capital availability, newer tools and technology enable entrepreneurs to do more with less capital. It's always a good time to become an entrepreneur if you see value that others do not. Prove the naysayers wrong by validating your idea in the market, which never lies.
Understanding cost implications of AI adoption for large vs. small enterprises
If you can make capex investments in AI today, it gives you a definite advantage in building AI systems fast with expensively running open source. Large language models (LLMs) with trillions of parameters are cheaper to deploy and use compared to small language models (SLMs), which require specialized workforce but are cheaper to run. The trade-off depends on capital availability: constrained capital leads to building SLMs, while unconstrained capital allows for faster cycle times with higher capex.
