Open Source Summit India 2025

India, US, EU, and other governments introduced cybersecurity regulations for anyone distributing software. Any software maintainer, contributor, and developer, needs to be aware of their software dependencies and any associated risk (vulnerabilities, licensing issues, unmaintained software, unknown origin), and how to efficiently manage these software components. This is most often – and now regulated – with SBOMs.

Container security approaches often focus heavily on image scanning while neglecting runtime protection, creating dangerous blind spots in cloud native architectures. Organizations deploy seemingly secure containers only to discover exploits targeting runtime vulnerabilities that bypass static analysis entirely. Our talk introduces a comprehensive defense-in-depth strategy that combines supply chain validation, admission controls, and runtime behavior analysis to detect and prevent sophisticated attacks. We'll demonstrate real-world attack patterns and how open source toolkits like Falco, SPIFFE, and AppArmor can be used for runtime protection, focusing on workload attestation and behavioral fingerprinting techniques that traditional security scanners often miss.

The growing complexity of cloud-native environments has made incident detection and remediation increasingly challenging, with engineers often spending precious hours sifting through disconnected telemetry data. This talk demonstrates how Model Context Protocol (MCP) which an open protocol for building agent applications bridges the gap between OpenTelemetry data and AI-powered analysis to dramatically reduce Mean Time to Recovery.

As enterprises race to harness GenAI's transformative power, they're caught in a critical dilemma: innovate rapidly or sacrifice control? Commercial AI Model Providers offer convenience but demand a steep price—your data sovereignty.

Generative AI has been very powerful in providing a natural language understanding and reasoning abilities to products and applications using approaches and technologies such as LLMs, RAG, vector databases and knowledge graphs. These capabilities have been significantly enhanced using orchestration and autonomous agents with frameworks such as LangChain and LangGraph. These powerful capabilities however bring in additional challenges when testing and evaluating the resulting products and applications because of natural language interface, non-deterministic output and additional complexity with the use of a hierarchy of autonomous agents.

As AI evolves from isolated model development to system-level deployments, enterprise AI engineers face increasing complexity in tooling, governance, and operational workflows. This session explores how open source technologies—Podman Desktop, RamaLama, and Llama Stack—can streamline the design and delivery of secure, scalable, and reproducible agentic AI systems. Attendees will follow an end-to-end user journey: starting from secure container-based local experimentation, advancing through modular multi-agent RAG workflows, and culminating in OCI-compliant model packaging and Kubernetes-native deployment. We’ll dive into practical strategies for integrating model governance via the LF AI & Data Model Openness Framework, achieving robust supply chain security, and enabling observability across the AI lifecycle. Whether you're building internal AI platforms or modernizing MLOps, this session reveals how open tooling empowers teams to confidently operationalize next-gen AI systems at enterprise scale.

AI systems are powerful, but too often they operate as black boxes—making decisions without any explanation. In high-stakes areas like healthcare, manufacturing, or compliance, that’s a real problem. This session introduces a simple but effective open framework for building explainable AI using modular models and domain-driven knowledge graphs.

Enterprise adoption of GenAI, to unlock its full potential is hindered by costs and a talent gap which needs to span AI and Cloud technologies. The Open Platform for Enterprise AI seeks to democratize AI, filling a talent gap and illustrating that one can deliver low latency edge solutions and cost-effective cloud solutions using a variety of hardware. Towards this, OPEA brings together state of the art models, containerized microservices for guardrails, vector databases, model servers and more, easily customizable blueprints for common use cases such as Chat Bots, Document Summarization, Agents to name a few, a no-code/low-code studio to create GenAI pipelines, cloud deployment and evaluation tooling. OPEA currently supports Intel, AMD, and Nvidia hardware and can be deployed in the cloud, on-prem, or on a shrink-wrapped appliance. It integrates with identity and authentication systems and includes tooling for monitoring and visualization, which enable delivering enterprise grade applications, accelerating transition from proof-of-concept to production. Come learn about OPEA, use it, influence it, make your mark on democratizing GenAI.

Multi-agent systems are revolutionizing industries, but their complexity introduces vulnerabilities from adversarial behaviors to privacy breaches that can undermine trust. How can we protect these systems without curtailing their transformative potential? In this session, we’ll explore best practices for identifying and mitigating unique security and safety risks, while preserving open source innovation. Through real-world examples, learn how to spot adversarial threats, prevent data leakage, and integrate robust governance structures that balance compliance and creativity. We’ll also demonstrate proven techniques and open source tools for safeguarding multi-agent systems, from model validation and continuous monitoring to ethical oversight and regulatory alignment. Whether you’re an AI developer, data scientist, or open source maintainer, you’ll gain concrete strategies to future-proof multi-agent AI, fostering an environment where innovation thrives without compromising security or trust.