Open Source Summit India 2025

India’s pursuit of Sovereign AI—the ability to build and run AI systems within national borders—is fast becoming a strategic imperative. This session explores how open source principles and infrastructure are accelerating India’s journey toward digital self-reliance, highlighting collaboration between local cloud providers and the open source community.

AI Agents are evolving from simple automation scripts to sophisticated decision-makers capable of reasoning, planning, and executing multi-step tasks across cloud environments. However, taking these agents from prototypes to reliable, observable, and scalable production systems remains a huge challenge.

Deploying one of the first multi-agent AI systems at scale across multiple European countries for Deutsche Telekom provided unique insights from vendor lock-in to ensuring regional sovereignty. This talk shares insights and real-world experiences from successfully creating and deploying an open, sovereign AI agent platform. By designing for cloud neutrality and decentralization from the ground up, we navigated complexities around compliance, scalability, and operational flexibility, turning potential barriers into strategic advantages. The project is now part of the Eclipse Foundation. https://eclipse.dev/lmos/ and part of several open groups in defining the future of Agentic systems like http://las-wg.org

The presentation explores the growing threat of quantum computers to traditional cryptographic algorithms used in embedded Linux systems. It begins with an overview of current security algorithms (RSA, ECC, AES) and their vulnerabilities to quantum attacks.

Supporting a device in a massive and ever-evolving project like AOSP isn’t just about getting it to boot—it’s about keeping it alive through a relentless churn of upstream changes. Whether it’s adapting to the latest AOSP updates, syncing with evolving Linux kernel versions, or integrating changes from external projects, the challenge is real—and constant.

This hands-on tutorial covers practical use of Rust both in kernel space and user space for building memory-safe, reliable systems using Linux eco system.

This talk would highlight the importance of open source health devices and how we used Zephyr across several hardware platforms to develop these devices. We will draw on experiences from our projects including HealthyPi 5, HealthyPi 6 and the wearable HealthyPi Move, all open source hardware and software. More specifically how Zephyr enabled the use of practically a single codebase across three different microcontroller platforms in different form factors.

This session will be about my current and ongoing experience at Texas Instruments where I have been working on enabling Zephyr for AM243x platform, a multiprocessor heterogeneous evaluation module with both Cortex-R and Cortex-M cores and all the challenges I have faced in the process. This includes adding board support, device trees, device drivers, as well as testing the peripherals to ensure proper function on Zephyr. I will also be walking through the process of writing a device driver and the device driver model for Zephyr, and my experience and the challenges I encountered in doing so. The talk will also include the gaps and challenges involved in booting a heterogeneous SoC, the limitations I faced with MCUBoot as a bootloader for Zephyr, the potential alternatives, and the solution involving TI's RTOS bootloader. The aforementioned topics may serve as a reference guide for enabling more heterogeneous SoCs in the future.

Twister, as Zephyr RTOS’s test runner, operates by automating the testing pipeline—discovering test cases, building them for target platforms, executing them on hardware or simulators, and reporting results. In this presentation will talk about how twister works like test case identification, configuration and build executing on hardware and how results are being collected