On May 18th, 2024, E2E Networks hosted AI Day Bangalore at Accel Launchpad. The event brought together AI experts to explore various facets of this rapidly evolving field. Three speakers, each with a distinct perspective, shared their insights on critical topics related to AI security, user experience, and model development.
.jpeg)
Next, Riyaz Walikar, co-founder and Head of Research at Kloudle, took the stage. His talk, titled "Hacking AI Applications on Kubernetes Clusters," addressed a critical aspect often overlooked in the AI development process – security. Riyaz, an expert in cloud and Kubernetes security, highlighted the growing prevalence of AI applications and the new security risks that come with them. His talk served as a wake-up call for developers and founders, emphasizing the importance of building security into AI applications from the ground up. He explored common vulnerabilities, potential attack scenarios, and offered practical solutions to keep AI systems safe. The highlight of Riyaz's talk was a live demo showcasing an attack in action, followed by actionable steps for developers and engineers to mitigate such risks.
How to Not Build Foundation Models by Abhishek Upperwal
The final speaker of the day was Abhishek Upperwal, a Data Scientist and CEO of Soket Labs. Abhishek concluded the event with a deep dive into the challenges and learnings associated with building large language models. His talk, "How to Not Build Foundation Models: Deep Dive into Learnings from Pragna-1b, Our Multilingual Open Source Language Model," offered valuable insights from the development of Pragna-1b, an open-source language model designed for Indian languages. Abhishek focused on the practical aspects of model development, sharing both successful strategies and missteps encountered during the training process. This talk provided valuable guidance for anyone interested in building their own AI models.
.jpeg)
AI Day Bangalore offered a well-rounded perspective on the current state of AI. The event covered critical aspects like security, user experience, and model development, highlighting the vast potential and challenges associated with this transformative technology.