Renumics

Interested in AI-driven assistants for industrial applications? Our #Industrial #AI #Canvas provides a structured framework for building Retrieval Augmented Generation (#RAG) systems tailored to industrial needs. Inspired by Alexander Osterwalder's canvas model for business innovation, the Industrial AI Canvas adapts it to RAG-based systems with key components: - Core: Value proposition, user interaction, and ML challenges - Data Science: Data sources, methods, and curation - Application: Integration, operations, and resources This Canvas bridges gaps between stakeholders, enabling seamless collaboration among domain experts, ML teams, and IT professionals.

The MLOps Community podcast was one of the podcasts I always enjoyed listening to. So, it was particularly exciting that Demetrios Brinkmann had asked me to contribute. In the latest Coffee Session episode, you can hear me talk about how data visualization and embeddings can support you in understanding your machine learning data. Here are some insights from our discussion: - Visualization Techniques: I discuss using UMAP to simplify data into two-dimensional maps to identify clusters and anomalies in large ML datasets. - Practical Applications: We explore how these techniques are applied in real-world scenarios, particularly in collaborations with automotive companies to enhance simulations and data analysis. - Challenges with Data: We address the challenges of large, unstructured datasets and the need for better tools. - Feedback and Iteration: We talk about the importance of incorporating real user feedback into the development of models and visualizations. 🎙️ Visualize - Bringing Structure to Unstructured Data // Markus Stoll // #258 https://lnkd.in/gyHpYr-V

Self-Supervised Audio Learning in the Hugging Face 🤗 ecosystem? I am currently working on a series of articles about the use of the Audio Spectrogram Transformer (AST) in industrial AI, and I have great news for you. The first article in the series, "How to use SSAST Model Weights in the HuggingFace Ecosystem?" is now live on the Hugging Face Community Blog! 🎉 🔊The article is a short tutorial on how to load the weights from the Self-Supervised AST model into the Hugging Face 🤗 Transformers model implementation for easier use and fine-tuning. 📝 Read the article here: https://lnkd.in/eES5jjd7 #SelfSupervisedLearning #AudioMachineLearning #HuggingFace #AI #MachineLearning #AudioSpectrogramTransformer #AST #Tutorial

For his debut TDS article, Marius Steger presents a detailed guide to fine-tuning the Audio Spectrogram Transformer model for audio classification of your own data, using the Hugging Face ecosystem.

I am happy to share that I was invited to a Podcast show to share some insights about industry use cases for AI and my career path to Renumics You can see me and Thomas Bustos talking about - Collaboration in projects and the Importance of feedback loops - Using Data visualization to find patterns and cluster in your ML Data - Data visualization for RAG - Our Open Source Visulization Tool Renumics Spotlight in the latest episode of "Let's Talk AI" at youtube: https://lnkd.in/eyvSiYhk

The Audio Spectrogram Transformer (AST) outperforms traditional CNN-based methods, but it needs significantly more data. Additionally, the use of a pretrained model trained on internet data may not be suitable for your #industrial #AI use cases. 📖 In this tutorial https://lnkd.in/eaU8Q9_K, we show you how to use the SSAST (Self-Supervised Audio Spectrogram Transformer) framework to train a model on your own unlabeled data and convert it into an easy-to-use Hugging Face AST. For the conversion, a conversion script is used that maps the SSAST layer names to the corresponding Hugging Face layer names.

🔌 Loading SSAST Model Weights into HuggingFace's AST implementation I'm pleased to announce my latest article on integrating SSAST model weights into Hugging Face's AST implementation, published in ITNEXT. This approach enables advanced audio classification capabilities using self-supervised learning and HuggingFace's powerful tools. 🔍 Why integrate? Leverage self-supervised learning for better feature extraction and model performance, and fine-tune the model for audio classification tasks with the convenience of the HuggingFace ecosystem. 🔧 How to do it? Step-by-step guide to seamlessly loading and fine-tuning SSAST weights using HuggingFace's tools. 🔗 https://lnkd.in/ei_KmZkX

Hey ML-Engineers, we are happy to share our latest project: A #RAG GUI with with Interactive Data Exploration (GitHub: https://lnkd.in/eCFsqH-c) You can use it to: -> Setup a RAG Pipeline for your local documents with a few command lines. -> Index your data (or download our prepared dataset) -> Use a GUI to ask questions and get answers together with the used document snippets. -> #visualize and interactively #explore: Most importantly, interactively visualize and explore all your asked questions together with the relevant document snippets in renumics-spotlight. The RAG System supports OpenAI Models like #GPT4 and also running local models from Hugging Face Models for #embedding generation and for the #LLM part. You can find more details and an example in our blog post: https://lnkd.in/eWN_Hdzz The animation below showcases the visualization approach with an increasing number of questions asked, resulting in more documents from the vector store being referenced by one, two, or more questions. --- Discover how Retrieval-Augmented Generation (RAG) is transforming industrial workflows in automotive, aerospace, and mechanical engineering: https://lnkd.in/enQdmdRK

Hey ML Engineers and Data Scientists, I'm excited to share my latest article: "Visualize your RAG Data — EDA for Retrieval-Augmented Generation." The article provides a step-by-step tutorial on how to create an interactive visualization for RAG embedding data with Renumics Spotlight (GitHub: https://lnkd.in/eURkgmb2). 🚀 You will build a visualization for a LangChain Retrieval-Augmented Generation Application with ChromaDB based on OpenAI's text-embedding-ada-002 and GPT-4. The demo data (🏎️ Formula One Dataset built from Wikipedia articles) can be easily changed to your own data. 📖 Check out the full article on ITNEXT: https://lnkd.in/g5s8XPFe The animation here shows the UMAP dimensionality reduction of the embeddings of document snippets, colored by their relevance to the question "Who built the Nürburgring?"

Hey ML-Engineers! We are proud to see the visualization for RAG topic with Renumics Spotlight (https://lnkd.in/eUF2qJKt) is getting more traction. Our latest article "Visualize your RAG Data — Evaluate your Retrieval-Augmented Generation System with Ragas" at Towards Data Science https://lnkd.in/e4gXgjUr was selected as TDS Editors's Pick and boosted as Top Story by Medium. In the article, you will learn - How to briefly build a RAG system with LangChain and OpenAI's GPT4 for Formula One -Generate questions and answers for Evaluation - Evaluate the RAG system with Ragas Metrics - Most importantly how to visualize the results with Renumics Spotlight and interpret the results Follow the 'Visualize Results' section to easily apply the visualization to your evaluation and data strategy — what insights do you uncover? --- Discover how Retrieval-Augmented Generation (RAG) is transforming industrial workflows in automotive, aerospace, and mechanical engineering: https://lnkd.in/enQdmdRK