Mozilla.ai

Yesterday, we hosted an engaging Discord session, "Building Lumigator: Evaluating LLMs Made Simple", with over 35 participants actively joining the conversation! 💡 Key insights from the event: - Developers are keen on tools that simplify evaluating LLMs across diverse data types—like text, images, and tabular data. - Topics like robustness against hallucinations and attacks highlighted the growing need for adaptable evaluation frameworks. - The enthusiasm and thoughtful questions reinforced the importance of transparency and user-friendliness in AI tools. A huge thanks to everyone who joined and contributed to the discussion 💙 This level of engagement motivates us to keep evolving Lumigator 🐊 and other tools to better serve the community. 💪 Want to get involved? Explore Lumigator on GitHub and help shape its future: https://lnkd.in/gB9fznYs Let’s keep building together! #AI #LLMs #OpenSource #MozillaAI #DeveloperTools

📢 Join Mozilla.ai engineers for a tech talk on Lumigator, the open-source tool for selecting the best LLMs 📢 Perfect for all Software Engineers without deep ML expertise, this session includes a feature walkthrough, real-world demo, and a preview of what's ahead before Lumigator's General Availability in early 2025! 🚀 #OpenSource #LLMs #AI #DeveloperTools #MozillaAI

The behavior of ML models is often affected by randomness at different levels: from the initialization of model parameters to the dataset split into training and evaluation. Thus, a model's predictions—like the answers LLMs provide—are potentially different every time it is run, which makes evaluating model performances more complex than it appears. This post by Davide Eynard is an introduction to hypothesis testing and how it can be used to compare results from different experiments (for instance, evaluations of different ML models) and draw conclusions that are supported by statistical evidence. It's powered by a marimo notebook that can run entirely in your browser so you're able to experiment and test it for yourself! *** Special thanks to Akshay Agrawal and Myles Scolnick building marimo and providing help and super early feedback; and to Vicki Boykis and Oleg Lavrovsky for testing the notebook and offering great suggestions on how to improve it.

This blog post by Davide Eynard from Mozilla.ai, with input from Vicki Boykis, explains the basics of hypothesis testing. It's powered by a marimo notebook that runs entirely in the browser — marimo's interactive elements and Pyodide support make it well-suited to education https://lnkd.in/eEt8ykte

As we continue to build Lumigator, our platform that helps developers choose the right large language model for their projects, we'd love to hear from you and ensure we're addressing the needs of the community! Please help us shape the future of Lumigator by answering our survey with insights on the infrastructure you use and the challenges you face in your AI development journey. It's focused on developers working with AI/LLM, completely anonymous and takes less than 5 minutes. Thank you!

Introducing Lumigator 🐊 In today’s fast-moving AI landscape, choosing the right large language model (LLM) for your project can feel like navigating a maze. With hundreds of models, each offering different capabilities, the process can be overwhelming. That’s why Mozilla.ai is developing Lumigator, a product designed to help developers confidently select the best LLM for their specific project. Read our launch publication by Juliana Araújo and subscribe to get further updates!

Introducing Lumigator 🐊 In today’s fast-moving AI landscape, choosing the right large language model (LLM) for your project can feel like navigating a maze. With hundreds of models, each offering different capabilities, the process can be overwhelming. That’s why Mozilla.ai is developing Lumigator, a product designed to help developers confidently select the best LLM for their specific project. Read our launch publication by Juliana Araújo and subscribe to get further updates! https://lnkd.in/d-zz_d2U

Another day, another model. However, the effectiveness of these AI systems largely depends on the quality of the data used for training and evaluation. This is where the concept of ground truth plays a crucial role, as it refers to the accurate, real-world data that acts as the gold standard for training AI models and assessing their performance. Sandra Antunes, Product Operations Manager at Mozilla.ai, provides today an extensive overview of the importance of ground truth data in AI applications. #AI #machinelearning #datasets

Wow 😮 HuggingChat is now natively integrated inside of Mozilla Firefox 🤯 To activate it: Go to Settings > Firefox Labs > AI Chatbot > Pick HuggingChat This is super convenient 🔥 hat/tip Mozilla.ai

🚀 Exciting Opportunity Alert! 🚀 We’re hiring a Machine Learning Engineer to join Mozilla.ai’s engineering team. If you’re passionate about building open source, scalable, trustworthy AI solutions, this remote role is for you! 🌍 Mozilla.ai is driving the AI revolution with an open-source approach. Join us in developing our model selection platform, Lumigator. 📍 Remote (Europe, East Coast USA, Canada) Ready to shape the future of AI with us? Apply now through the link below!