SAP Inside Track Bangalore 2024: Generative AI Extravaganza

[vedant_gupta]

INTRODUCTION

It has been less than two months into the year, and we already have had a banger of an event! With over 1500 attendees, 56 mouthwatering technical talks and bustling conversations all around, we conducted the first edition of SAP Inside Track Bangalore 2024. This event has been shaping up to be a highlight of the year for SAP enthusiasts and tech aficionados alike for quite some time now. It started with an insightful and uplifting keynote by SAP Labs India MD and Head User Enablement, Sindhu Gangadharan where she talked about the impressive accomplishments by SAP and setting the tone for 2024.

Attendees then dispersed to explore the eight parallel tracks covering a diverse range of topics including SAP BTP, App Development, Data Analytics, AI & Automation, UI UX, SAP Concur, and S/4 HANA. Among these, Business AI took center stage, with Generative AI emerging as a focal point of discussion. The buzz around Generative AI was palpable as experts delved into its potential to streamline and optimize business processes. From brainstorming sessions to practical demonstrations, participants eagerly exchanged ideas on harnessing Generative AI to drive innovation and efficiency in their respective fields.

SESSION ON GEN AI REFERENCE ARCHITECTURE FOR BTP

At SIT, our session titled "Developing Gen AI Applications on SAP BTP following Reference Architecture" was a deep dive into the exciting world of generative AI and its practical applications within the SAP ecosystem. We began by laying the groundwork, exploring the basics of GenAI and elucidating the crucial distinctions between discriminative ML and GenAI. This foundational understanding is pivotal for effective app development, enabling us to brainstorm and devise strategies for harnessing the power of generative AI in our applications and also make efforts well directed. We also discussed the two main concepts that make Foundation Models and Large Language Modes (LLMs) so successful, token embeddings and transformer learning.

Moving forward, we delved into techniques such as Prompt Engineering, Retrieval Augmented Generation (RAG), and Fine-tuning for LLMs. With a particular emphasis on RAG, we unveiled its tremendous potential for addressing real-world business challenges and enhancing decision-making processes.

Following this, my teammate @praveenpadegal shed light on the Gen AI offerings from SAP BTP, including Joule, SAP Generative AI Hub, and SAP HANA Cloud Vector Engine. These offerings serve as invaluable tools for keeping our data, processes, and LLM operations within the BTP landscape, ensuring seamless integration and scalability. Our discussion then transitioned to real-world use cases that we've developed.

GEN AI USE CASES

We talked about the GenAI Mail Insights application, a multi-lingual support environment inundated with daily emails. Leveraging LLMs the application intelligently sifted through these emails, extracting pertinent information, determining urgency and sentiment, and preparing tailored responses. What set this application apart was its ability to provide translation capabilities. Using RAG the application can refer to previously received mails and accelerate the process of curating responses based on replies to the same. We took an example of a travel agency scenario. Here, the challenge lay in managing a high volume of emails in various languages, ranging from booking inquiries to itinerary changes. By applying the GenAI Mail Insights use case to this scenario, we showcased the immense value added by automating the entire email handling process. You can access the DC mission for this use case here. 

We then delved into the GenAI extension to our Events-to-Business framework, a pivotal enhancement aimed at addressing the significant challenges faced by users in API selection and usage. Built on the SAP BTP, E2B operates on an event-driven architecture, seamlessly integrating events from diverse non-SAP sources into SAP business processes to enhance efficiency. However, users often grapple with the time-consuming task of identifying the correct APIs for their specific use cases, consuming a considerable portion of development time. This complexity is compounded by the need to understand the APIs' capabilities, interfaces, and interactions with other systems, often necessitating extensive documentation review and technical comprehension. With the incorporation of RAG into our framework, we introduced built-in support and decision-making tools to guide users in API selection and usage. These decisions often come as chains of pre, main and post actions which work in sync. The GenAI extension can create the chaining for these action components to streamline the entire process. You can read about the E2B framework from this blog series. A deep dive technical article on the GenAI extension can be found here.

In the Q&A session that followed, everyone had the chance to ask questions and share thoughts. We discussed topics like how RAG works and how it can help businesses. People also wanted to know more about the tools offered by SAP BTP and how they could improve their work. It was a great opportunity for everyone to learn from each other and exchange ideas, making the event even more valuable for everyone involved.

OTHER SESSIONS BY MY TEAM - PLATFORM ADOPTION & ARCHITECTURE

My BTP guru colleagues gave some insightful sessions on various topics clutching more audience traction.

@lalitmohan  and @navyakhurana spoke on the topic "Federating Machine Learning (FedML) Reference Architecture for Hyperscalers with SAP Datasphere" where they discussed SAP's FedML libraries which are game-changers for businesses looking to harness the power of machine learning without the hassle of data extraction and migration. These Python libraries leverage SAP Datasphere's Data Federation architecture to intelligently source both SAP and non-SAP data for ML experiments, right at the hyperscaler's doorstep. They gave demos for using FedML with Databricks and Google Cloud Platform.

@Ajit_K_Panda and @Aryan_Raj_Sinha gave a house-full session on "Build Resilient Applications on SAP BTP" where they discussed the concepts from ground up and engaged in a deep technical discourse on building S/4 HANA extensions using patterns of DevOps, observability, integration with business process workflows and multi-region failover.

@uma-anbazhagan and @swati_maste gave a session on "Implementing End-to-End Use Cases with SAP BTP and Hyperscalers" discussing the synergy relationship between SAP BTP and other hyperscalers like AWS, Azure and Google Cloud Platform, discussing concrete end-to-end use cases and SAP's multi-cloud strategy.

It is always great to learn and give back to the community and SIT offers the best possible platform to do so. It was an exceptional event and I'm sure all attendees and participants had a ball! Onto more such interesting events.