Day 1

Pixie is a framework designed to improve digital marketing through a context-aware, multi-agent large language model (LLM). It supports real-time and offline marketing, using multimodal LLMs for enhanced reliability and performance. Pixie features a primary orchestration agent that coordinates workflows and task-specific agents that perform specialized tasks. By combining various learning techniques and marketing insights, the framework enhances decision-making and operational efficiency. It was tested for accuracy, consistency, and domain-specific problem-solving, showing improved performance in tasks like root cause analysis, campaign planning, and reporting.

Large Language Models (LLMs), trained on extensive internet-scale datasets, demonstrate exceptional performance across diverse tasks. Yet, they often fall short when it comes to grasping the unique nuances of organizational data. As a result, standard LLMs frequently require human intervention to deliver accurate, business-specific insights.

Join us for this Tech Talk as we dive into the world of LLMs. Explore the challenges LLMs face with business-specific contexts and how we can make them more effective. Discover strategies for integrating organizational context into models, examine how LLMs utilize this enhanced information, and uncover its transformative impact on the accuracy and specificity of their responses.

With AI applications, infrastructure, and tools getting rolled out at scale, it is important that we focus on the security and privacy around building, deploying, using and governing these applications. My talk will focus on both the technical information security aspects- (vulnerabilities in AI applications and associated tech stacks, treatment/mitigation strategies, proactive, detective and reactive security engineering aspects) and also from the GRC- governance, risk, compliance/legal aspects of both the security and privacy elements.

This research work tackles the challenge of accurately categorizing customers for retailers to enable precise, targeted marketing strategies. It introduces an innovative method leveraging Large Language Models (LLMs) to analyze purchase data and identify key product categories, such as baby food or eldercare items, revealing household compositions like families with kids or elders. By integrating historical behavior with LLM-based classification, the approach enhances life stage segmentation, improving customer targeting, sales, and retention. The research paper details the data sources, model architecture, and evaluation metrics used in this segmentation.

Session will help you define the Platform features for a successful implementation of Ops for Data and identify the challenges

In a world where over 90% of internet users speak languages other than English, the potential for growth lies in inclusivity. This talk will explore how AI is bridging the gap for over a billion users by transforming content into native languages at scale. I Himanshu Sharma, Co- Founder of Devnagri, will share insights on how neural machine translation, voice bots, and foundational language models are reshaping industries—creating opportunities for businesses to connect deeply with regional audiences. Dive into the future of AI-driven localization and learn how innovation is turning barriers into bridges for global communication.

Edge computing is revolutionizing earth observation (EO) by enabling satellites to process data directly in orbit, reducing reliance on ground stations for time-sensitive applications. This talk will highlight the benefits of edge computing for earth observation, such as reduced data transfer costs, faster response times in critical scenarios like disaster management, and real-time insights for sectors like agriculture and defence.

In September 2023 KaleidEO demonstrated on-orbit processing on EO satellites, showcasing real-time analytics capabilities like change detection and applications for vegetation, forestry and environmental monitoring. By enabling instantaneous data processing, edge computing ensures that actionable intelligence reaches decision-makers without delay, fostering more effective responses to natural disasters, better resource management, and stronger national security.

The edge computing innovation for EO marks an important evolution in the traditional approach, unlocking unprecedented potential for near real-time, data-driven, diversified applications from space.