Tool Name: CodeBaseBuddy with Codestral Key Features: - Privacy-preserving: Locally deployable, ensuring all operations remain on the user's machine. - Semantic Code Search: Utilizes advanced techniques for precise and relevant code search results. - Accelerated Onboarding: Quickly brings new contributors up to speed with any codebase. - Error Reduction: Provides specific guidance to minimize mistakes and prevent bugs. - Enhanced Engagement: Encourages contributions from those hesitant due to unfamiliarity with the codebase. - Continuous Learning: Helps both newcomers and experienced developers discover and learn about lesser-known parts of the codebase. - Technical Implementation: - **Vector Indexing**: Builds an annoy vector index by generating embeddings for every file in the repository. - **Query Processing**: Utilizes Annoy for efficient querying of the vector index. - **Guidance Generation**: Leverages locally deployed Codestral to provide step-by-step instructions based on user queries. - Original Contributions: - Local LLMs for Semantic Code Search: Uses Ollama's local large language models for running and managing models on a user's laptop. - Advantages:: - Privacy: Everything is deployed locally, ensuring user data remains secure. - Speed: Accelerates the process of getting acquainted with a new codebase. - Accuracy: Reduces the likelihood of errors with precise guidance. - Encouragement: Motivates more community contributions by simplifying the onboarding process. - Learning: Continues to be a resource for experienced developers to explore and understand different parts of the codebase. Summary: CodeBaseBuddy with Codestral is a cutting-edge tool designed to streamline the onboarding process for new developers, provide precise code search capabilities, and enhance overall engagement and learning within development teams. By maintaining a local deployment, it ensures privacy while offering robust and innovative features.

The Langflow Hackathon, led by Raghavan Muthuregunathan, addresses the challenge of enhancing LLM performance in low-resource languages. While some LLMs like GPT4o excel, others like Stable Diffusion struggle. The proposed solution, Translation Augmented Generation, leverages high-quality English LLMs for non-English languages using Langflow, Streamlit, and OpenAI. Their innovative prompt engineering allows Stable Diffusion to understand languages other than English, demonstrated by generating images from Tamil prompts. This solution aims to democratize LLM accessibility, making it beneficial for low-resource languages. More details and examples can be found on their [Streamlit app](https://langflow-lablab.streamlit.app/) and [GitHub repository](https://github.com/raghavan1988/langflow_hackathon/blob/main/app.py).

Cricket boasts over one billion fans in South Asia. - **Innovative App Features**: - **Enhanced Viewing Experience**: - Live commentary in the voices of famous personalities. - Engaging and entertaining match insights from legendary commentators, celebrities, and cricket icons. - **Coaching Tool**: - Helps coaches develop effective strategies against opponents. - Analyzes historical data, player performance, and match conditions. - Provides actionable insights and tactical recommendations. - **Benefits**: - Enriches the fan experience by making every match more engaging. - Empowers coaches to make informed decisions. - Enhances the overall competitiveness and enjoyment of cricket.

We are utilizing GPT-4o to develop a versatile search engine capable of understanding any language and answering any question. This advanced AI-driven platform leverages the power of natural language processing to provide accurate and relevant information across various topics. Our goal is to create a tool that transcends language barriers, making information accessible to everyone, regardless of their native language. Whether it's complex technical queries or everyday questions, our search engine aims to deliver precise answers quickly and efficiently, enhancing the user experience and knowledge accessibility.

Challenge: Image generation from text in local languages. Tracks: helps all 3 categories of culture preservation, sustainable agriculture and promoting tourism The solution developed for the Soro Benin Hackathon aims to preserve local languages and cultural identity in Benin through AI-generated images and text. The application uses a prompting technique called Translation Augmented Generation to generate culturally nuanced images from prompts in Yoruba, Fon, and Dendi languages. It also answers questions related to agriculture, tourism, and current events in these languages. The solution leverages existing language models and APIs, saving costs compared to training new models from scratch. It addresses challenges in cultural preservation, sustainable agriculture, and tourism by providing AI visuals and information in local languages, aiding communication and showcasing Benin's unique heritage to drive sustainable practices.

Cricket VIDEO ANALYZER to understand a batsman using 12 labs API +llama index + GPT 4 Solution: Choose 1 batsman (Glenn Maxwell of Australia) (Source 1) : Use 12 labs API to understand their playing style - Understand the video Generate the text (Source 2 for additional augmentation) : Use cricinfo.com APIs to get ball by ball commentary of the batsman Build a vector store of the dataset built using (2) and (3) Use GPT-4 to answer questions on this dataset (RAG application) Global Popularity and Demand: The immense popularity of cricket has led to a high demand for video analysis tools. Coaches, players, and teams use these tools to analyze player performance, match strategies, and opponent weaknesses. The ability to break down every aspect of a game, from batting techniques to bowling actions, has become crucial for success. As cricket leagues (such as the Indian Premier League) and international tournaments continue to grow, the demand for sophisticated video analysis software has skyrocketed. Advanced Technology and Data Insights: Cricket video analysis tools have evolved significantly over the years. High-speed cameras, ball-tracking systems, and AI algorithms allow for precise analysis of player movements, ball trajectories, and field placements. Coaches can identify patterns, assess player fitness, and make data-driven decisions. For example, analyzing a batsman’s scoring zones or a bowler’s release point can provide valuable insights.

Chat with your PDF is the standard RAG 101 example. In this hack, we demonstrate an example of Chat with your code base. Problem: Imagine you are trying to contribute to a new code base (a github repository) for a beginner task. Knowing which file to change, where to make the change can be time consuming. This is NOT about CODE COMPLETION. This a semantic code search application deployable locally and works from a terminal. Business impact: Developers would have enhanced productivity and be willing to buy this RAG application Code completion tools do not tell where to make the changes on an existing code base.

Problem Statement: 1) Over 70% of PDFs contain critical data in images like charts and tables, especially research articles 2) Gemini is released for English only today. Can we build a solution for 1) Answering natural language questions based on images in PDFs ? 2) Making Gemini accessible for non english speakers? By leveraging Spire, Open AI GPT 3.5, Gemini Pro Vision and Trulens, I have built an application that solves both problems - Spire for Image Extraction - Open AI for Translation to English (optional) - Gemini-Pro-Vision for the answer - TruLens for Monitoring

100s of new arxiv papers are uploaded everyday How do I get new ideas to apply them for my field ? How do i get a quick gist of the paper and related papers? Loom video of Demo: https://www.loom.com/share/a9d151a59a7c4fe68a8cc94b6a118d9d?sid=90a5d924-d995-4716-8a3b-c4da861e2e1d This helps improve creativity of researchers by applying the ideas of a specific ARXIV paper to a completely new field. This helps create a gist of the large arxiv paper with just the URL There by improving their productivity It also helps create a summary of all the related papers by doing a vector search based on the title

🗓️ This will be a 7-day virtual hackathon from 13-20 January 💻 Access AI21 Labs' state-of-the-art language models to build innovative applications 🏆 Prizes and awards of up to $9000 API credits + $3500 cash 💡 Meet and learn from AI21 Labs and Lablab experts 👥 Join the community and find your team ✔️ All levels are welcome 📖 AI tutorials and mentors to help you ⭐ Receive a certificate of completion and Wordtune premium for submission 🐱‍💻 Sign up now! It’s free!

🗓️ This will be a 7-day virtual hackathon from 16-23 December 💻 Build AI application with the latest large language model-powered technology by Cohere 💡 Get the chance to work with the best AI professionals in the industry and learn from them ✔️Entry level = 0. You’ve just started with AI? Are you an experienced Data Scientist? Or maybe you are a Designer or Business Developer? Join us! We need your domain knowledge! 🐱‍💻 Register now and let's get started! It’s free!

🚀💻 Be the first to build an AI App on Google's models! Hackathon on July 7-10. 🔬🌐 Try new Vertex AI features from Google Cloud Platform. 🤝🌍 Learn from AI leaders and connect with like-minded people. 🛠️📱 Build apps with the world's best AI tools! 💡🌍 Solve real-world problems with Generative AI models