From a "simple" finetuning to your own Mixture of Expert model using opensource models. Nowadays training from scratch an LLM is a so huge effort also for very big company. Starting from pre-trained models to create your own model is no more a way for resourceless companies, but a often a must starting point.
- Lora - Quantization and QLora - Injecting embeddings model into Lora to manage multiple Lora adapters. - Mixing models - Creating your MoE (Mixture of experts) model using several finetuned (Your own) models
Winner of three AI awards, I’ve been working in AI and machine learning for 25 years, designing and developing AI and computer graphic algorithms.I’m very passionate about AI, focusing on Audio, Image and Natural Language Processing, and predictive analysis as well.I received... Read More →
Anil Inamdar, Instaclustr by NetApp, Head of Data Solutions
For dev teams building and training their own AI models, ending up with AI solutions plagued by hallucinations and reliability issues is a (rightfully) huge concern. The good news: vector databases make generative AI considerably more reliable and less prone to hallucinations. The even better news: a number of *100% free and open source* vector databases are especially great options for supporting AI workloads.
The good news keeps coming for teams considering an open source vector database path: it isn’t necessary to invest in implementing new or exotic or proprietary or specialized data-layer solutions to harness vector databases. Many enterprises will find that their existing infrastructure can already support AI workloads (while continuing to provide the familiar data availability, scalability, and performance they already know they can trust). In particular, PostgreSQL (with the pgvector extension), OpenSearch, and Apache Cassandra 5.0 (with its new native vector indexing) are three completely open source technologies—no proprietary or open core solutions needed—that tick all the boxes for meeting enterprises’ AI workloads requirements.
Attendees of this Dev Innovation Summit session will learn how open source vector databases utilize vector embeddings to enable more accurate LLMs, and how to strategically approach and implement retrieval augmented generation (RAG) processes. Attendees will also come away with a clear understanding of the advantages offered by PostgreSQL with pgvector, OpenSearch and Cassandra 5.0 as particularly mature open source vector database strategies ready to jumpstart your organization’s competitive AI capabilities.
Anil Inamdar is the VP & Head of Data Solutions at Instaclustr by NetApp. Anil has 20+ years of experience in data and analytics roles. Joining Instaclustr in 2019, he works with organizations to drive successful data-centric digital transformations via the right cultural, operational... Read More →
Ankit Jain, Aviator Technologies, Co-founder & CEO
In this talk we will delve into the nuanced world of software engineering, where the emphasis on metrics can often overshadow the crucial aspect of developer experience. We will explore how an over-reliance on metrics can hinder innovation, cause unexpected behaviors, and ultimately erode team morale.
Discover alternative approaches to scaling developer experience that prioritize human-centric methodologies over mere data points. We will also dive into practical strategies to use automation and ownership to scale developer experience.
Ankit is a cofounder and CEO of dev-productivity startup Aviator Technologies and also leads the ex-Google alumni network (Xoogler.co). Previously he led engineering teams at Sunshine, Homejoy and Shippo. Prior to that, Ankit was also an EIR at Unshackled Ventures and an engineer... Read More →
Gaganjot Kaur Kang, Sony PlayStation, Senior Software Engineer
This session is designed to provide a comprehensive framework for developing real-time data analytics and data processing pipelines and to explore the cloud-based technologies that facilitate this process like AWS Timestream, and AWS S3. The agenda includes:
- An overview of the general architecture of real-time event processing systems - The role and importance of technologies such as Apache Kafka, Apache Flink, AWS Timestream, AWS S3, and others in this architecture - Addressing challenges in data processing, including state management, event-time handling, and job monitoring - Real-world applications, such as metric aggregation over a time window and real-time computation of product prices and discounts
