LLMs and VLMs - The Monthly Dev #47

**The Monthly Dev brings world-class speakers to empower the developers' community, once a month. Made with ❤️ by daily.dev.**

Agenda:
We're excited to announce our next event in this series with an awesome lineup of speakers! See details below.

Opening remarks by Francesco Ciulla (Developer Advocate at daily.dev)
https://twitter.com/FrancescoCiull4

Talk 1

Sumanth Papareddy, ML Dev Advocate, Clarifai

https://x.com/Sumanth_077

Talk title: Fine-Tuning LLMs: A No-Code Approach with Clarifai Platform

Abstract:
Abstract: While techniques like Prompt Engineering and Retrieval-Augmented Generation (RAG) can address many limitations of Large Language Models (LLMs), fine-tuning provides a more targeted solution when these methods fall short. In this session, we will examine the limitations of Prompt Engineering and RAG, and dive deep into the process of fine-tuning to tailor LLMs for specific tasks or domains using the Clarifai Platform. We'll explore how to refine pre-trained models, such as the latest Open Sourced Llama 3.1 model, by training them on specialized datasets to enhance their performance.

Talk 2

Amir Alush, co-founder and CTO of Visual Layer

https://x.com/alushamir

Talk title: From Raw Data to Refined Datasets for AI Model Development

Abstract:
The success of AI models relies heavily on high-quality, extensive image datasets for effective training. Whether you're an AI enthusiast, researcher, or student, you've likely faced challenges with messy image datasets. Issues like mislabeled data, duplicated images and outliers can severely impact model performance, waste computational resources and storage, and require significant manual work which is costly and inaccurate.
Building on the success of our Python project fastdup, which has garnered over 400K installations, we've introduced the Visual Layer cloud platform. This platform empowers users to organize, explore, enrich, and extract visual datasets (images and videos) at scale, accessible from anywhere.

Talk 3

Graham Miller, Fullstack Developer at Private AI

Talk title: Reflections On Building Our First LLM-Based Application

Abstract: Building and shipping a product that people actually like using is a challenging task. At the beginning of the GenAI hype-cycle early last year, PrivateAI built, deployed, and maintained a privacy preserving LLM based chat app called PrivateGPT. PrivateGPT since launch has had approximately 10,000 users.

In this talk, I will take you on a journey through the initial days of product ideation, to prototyping and building PrivateGPT, and finishing at the current state of the product. I will bring you through the decisions we made during those early uncertain periods of product development - the ones that proved beneficial and those that turned into learning experiences for our next product.

Starting time: November 26th, 8AM PT (SF) / 11AM (NYC)