AI Makerspace by Digital Product School

🚀 Mixture-of-Agents (MoA) combines the strengths of multiple LLMs. It serves a hierarchical architecture of agents to optimize performance and cost efficiency. The #opensource AI community join forces to shape the future of MoA. 🔍 Key Highlights: ✨ Ensemble of Models: LLMs generate better responses by collaborating and refining outputs. ✨ Layered Structure: Multiple layers of LLMs iterate to produce high-quality responses. ✨ Top Performance: MoA outperforms models like GPT-4 Omni on benchmarks in several dimensions. 💡 Practical Benefits: ✨ Improved response quality. ✨ Reduced retraining costs. ✨ Scalable across various LLM architectures. The MoA methodology offers a scalable, cost-effective solution for enhancing LLM capabilities. Thanks to the development team from Duke University, Together AI, University of Chicago and Stanford University. Check out the links for a quick start and demo: 📜 Blog: https://lnkd.in/eFwkqk6N 👩💻 CLI Demo: https://lnkd.in/gifKtUQv Stay updated with the latest in AI tools! Follow AI Makerspace by Digital Product School for more insights. 🤖 #NLP #AI #MachineLearning #Innovation #LLM #MoA

🚀 Exciting News in AI Development: 𝐄𝐧𝐭𝐞𝐫 𝐭𝐡𝐞 𝐄𝐫𝐚 𝐨𝐟 𝐆𝐫𝐚𝐩𝐡𝐑𝐀𝐆! 🌟 GraphRAG merges retrieval-augmented generation (RAG) and query-focused summarization (QFS) to enable LLMs to answer global questions over private and unseen document collections. It builds a graph-based text index in two stages: deriving an entity knowledge graph from the source documents and pregenerating community summaries for closely-related entities, ensuring comprehensive and accurate answers. 💡 Here's what makes GraphRAG stand out: 🔍 𝐄𝐧𝐡𝐚𝐧𝐜𝐞𝐝 𝐀𝐜𝐜𝐮𝐫𝐚𝐜𝐲: Combining vectors and knowledge graphs, GraphRAG delivers up to 3x more accurate responses, significantly improving the quality of answers. 💻 𝐃𝐚𝐭𝐚 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲: With up to 97% reduction in token usage, GraphRAG ensures scalability and cost-effectiveness, making it ideal for large-scale applications. 📊 𝐄𝐱𝐩𝐥𝐚𝐢𝐧𝐚𝐛𝐢𝐥𝐢𝐭𝐲: Knowledge graphs offer a visual and intuitive way to explore and understand data, providing unparalleled insight and facilitating faster iteration and development. 🛠️ But GraphRAG doesn't stop there! Here's how it empowers researchers and developers: 🎯 𝐁𝐞𝐭𝐭𝐞𝐫 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞: Improved explainability and traceability enhance data governance, crucial for industries like finance and healthcare. 🌱 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭 𝐄𝐱𝐩𝐥𝐨𝐫𝐚𝐭𝐢𝐨𝐧: Visualize and debug data effortlessly with knowledge graphs, accelerating research and innovation. 🏗️ 𝐑𝐨𝐛𝐮𝐬𝐭 𝐏𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞: Integrate GraphRAG into your applications to achieve superior performance and scalability. 🔗 Ready to dive in? Start here: https://lnkd.in/gDM4AJYh Stay updated with the latest in AI tools! Follow AI Makerspace by Digital Product School for more insights. 🤖 #AI #GraphRAG #GenAI #MachineLearning #KnowledgeGraphs #Innovation #Microsoft #Neo4j

🎬 Introducing the Future of Scalable AI Workflows! 🌟 LangChain has just announced their latest breakthrough: 𝑳𝒂𝒏𝒈𝑮𝒓𝒂𝒑𝒉 𝑪𝒍𝒐𝒖𝒅. Here's why this is a game-changer: • 𝐑𝐞𝐥𝐢𝐚𝐛𝐥𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐒𝐲𝐬𝐭𝐞𝐦𝐬: LangGraph Cloud provides infrastructure for deploying your LangGraph agents at scale. Enjoy seamless fault tolerance and scalability for robust AI applications. • 𝐈𝐧𝐭𝐞𝐠𝐫𝐚𝐭𝐞𝐝 𝐌𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠: With LangGraph Studio, visualize agent trajectories, debug failure modes, and iterate rapidly. Perfect for prototyping, debugging, and monitoring complex workflows. • 𝐄𝐚𝐬𝐲 𝐃𝐞𝐩𝐥𝐨𝐲𝐦𝐞𝐧𝐭: Deploy your agentic application with a single click from a GitHub repo. No infrastructure expertise required, simplifying the process from development to production. • 𝐅𝐞𝐚𝐭𝐮𝐫𝐞-𝐑𝐢𝐜𝐡 𝐄𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞: LangGraph Cloud supports streaming, human-in-the-loop interactions, double texting strategies, background jobs, and cron jobs. Enhance your AI systems with these advanced capabilities. • 𝐇𝐮𝐦𝐚𝐧-𝐀𝐠𝐞𝐧𝐭 𝐂𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐨𝐧: Design agents that can wait for human approval, allowing for real-time inspection, editing, and resumption of workflows, ensuring high-quality output and reliability. Join us in exploring the future of AI with LangGraph Cloud. The revolution in AI scalability and reliability is here. 🌟 Follow AI Makerspace by Digital Product School for the latest updates and insights into AI technology! 🚀🤖 #AI #MachineLearning #LangGraphCloud #LangChain #AIInnovation #AgenticSystems #FutureTech #TechInnovation #ScalableAI

We're excited to bring updates to four new amazing AI tools that are set to change the game. Check them out! 👇 🔊 𝐒𝐨𝐮𝐧𝐝𝐞𝐟𝐟: Elevate your audio production with AI-driven sound effects! • 🎧 Convert text into custom, high-fidelity sound effects with ease. • 🌟 Ideal for crafting unique ambient noises and specialized sounds. • 💼 Includes commercial usage rights for professional projects. 📊 𝐏𝐫𝐨𝐦𝐞𝐭𝐀𝐈: Revolutionize your business planning process! • 📝 Develop detailed business plans and valuations seamlessly. • 📈 Integrates strategic tools like SWOT Analysis, PESTEL, and more. • 🔒 Prioritizes data security and intellectual property protection. 🏠 𝐍𝐞𝐨𝐥𝐨𝐜𝐮𝐬: Transform home interior design with AI precision! • 🖼️ Generate high-quality renders of your space in under a minute. • 🛋️ Provides product recommendations to actualize your designs. • 🌍 Showcased at prominent events like the ROSHN PropTech Hackathon. 🖼️ 𝐃𝐫𝐚𝐠 𝐘𝐨𝐮𝐫 𝐆𝐀𝐍: Unparalleled control in visual content generation! • 🎨 Manipulate pose, shape, expression, and layout of generated objects. • ✨ Interactively 'drag' image points to precise target positions. • 🖌️ Handles diverse categories, from animals to landscapes, with exceptional precision. Harness the power of these AI innovations to elevate your projects to new heights! ✨ 🔔 Follow AI Makerspace by Digital Product School for the latest updates and insights on groundbreaking AI tools! 🚀 #AI #AI_Makerspace #news #innovations

🚀 #𝐒𝐮𝐜𝐜𝐞𝐬𝐬𝐒𝐭𝐨𝐫𝐲: Team Avishkar AI of Digital Product School by UnternehmerTUM developed 🤖 ai-SAP in partnership with SAP, which provides one-point access to information spread across Slack, GitHub and Internal Documentation ASAP. It enables the user to search the knowledge base, ask questions, generate code and have conversations with it, saving hours (and sometimes days) spent on Stack Overflow, Slack, GPT, GitHub and scattered documentation while debugging. 🤝 𝐔𝐬𝐞𝐫 𝐏𝐫𝐨𝐛𝐥𝐞𝐦 Scattered knowledge across sources makes it very time-consuming and tedious to search for the right information and a keyword-based search is often not useful enough. This makes the developers search for solutions on multiple platforms or reach out to colleagues which consumes hours of their valuable time. 💡 𝐒𝐨𝐥𝐮𝐭𝐢𝐨𝐧 𝐈𝐝𝐞𝐚 • ai-SAP enables the user to: • Search for information in natural language. • Ask questions and have conversations with the knowledge base. • Go to most relevant linked sources extracted from the knowledge base. • Give feedback about the conversation. 🎯 𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬 𝐨𝐟 𝐀𝐈 • Data loaders in 15 formats for the knowledge base. • Retrieval-Augmented Generation to extract most relevant sources, answer and code. • Query and Chat Engines to enable conversations and regeneration of answers. • Generation-Reflection cycle for the LLM to satisfy user requirements. • Feedback loop to augment knowledge base from negative sentiment conversations, and enable continual learning. 💻 𝐓𝐨𝐨𝐥𝐬 𝐔𝐬𝐞𝐝: #LlamaIndex #TogetherAI #Ollama #Docker #GCP #Redis #RAG 🤗 𝐓𝐞𝐚𝐦 𝐦𝐞𝐦𝐛𝐞𝐫𝐬 Pulkit Saini - Product Manager Alexandra Avramova - Interaction Designer Diksha Shrivastava - AI Engineer Tony LE - Software Engineer Adrian Regenfuß - Software Engineer 💎 𝐓𝐞𝐬𝐭𝐢𝐦𝐨𝐧𝐲 𝐛𝐲 Diksha Shrivastava “DPS has been the brightest career experience for me till now. I joined DPS with the expectation of being challenged, growing and developing something meaningful, and the experience completely exceeded my expectations. Working with a cross-functional team in a very diverse atmosphere, I felt empowered creatively by the environment and by my team to create a valuable product and to keep improving in iterations. DPS pushed me to be brave and bold, taught me to follow the simplicity principle and instilled in me a strong desire to come back some day with my own product. If you want to engage in product development, work with cutting-edge technology and create some meaningful impact, I highly recommend you to join DPS.” #genAI #RAG #LLM #AI #DigitalProductSchool

🚀 #SuccessStory: Team DAYOM from Digital Product School by UnternehmerTUM developed 🐸 RENE, a negotiation assistant to streamline supply-chain negotiations. RENE empowers purchasing officers with AI-driven, up-to-date procurement insights, enabling well-informed supplier negotiations. This innovation happened in collaboration with MAN Truck & Bus Deutschland GmbH. 🤝 #UserProblem: Procurement prices fluctuate due to economic trends, regulations, supply disruptions, currency changes, and market demand shifts. This forces purchasing officers to conduct extensive research, often taking over two weeks. 💡 #Solution: RENE reduces the extensive research time for purchasing officers from weeks to just a few minutes by enabling them to: ✔️ Select the supplier name, product and location of the good they are responsible for. ✔️ Generate price drivers for the goods. ✔️ Leverage LLMs to generate well-informed arguments based on recent, reliable news articles. ✔️ Lead successful renegotiations through an intuitive and simple process. 🎯 #Applications_of_AI: ✔️ Data collection: News article search and #ranking. ✔️ Break down the price-driving factors of the product. ✔️ Enhance news article search through query tuning. ✔️ Negotiation insights generation using a #RAG pipeline. ✔️ Provide arguments for negotiations supported by referenced articles. ✔️ Conduct financial #sentiment analysis for arguments related to price increases or decreases. 👩💻 #ToolsUsed: #LlamaIndex #LangChain #OpenAI #Docker #FastAPI #Qdrant #GPT #RAG 🤗 Meet the team members behind RENE Product Manager: Yallitza Loncharich Interaction Designer: Ananya Singh Software Engineers: Mauricio Erazo, Ole Vester AI Engineer: David Pop 💎 Testimony by David Pop: “DPS has been a great experience, giving us the chance to ideate and work together in an interdisciplinary team to solve real-world challenges of industry partners. As an AI engineer, you have the opportunity to work on impactful solutions based on genuine user needs. One of the most valuable lessons I have learned at DPS is taking into account the user value proposition for every engineering decision. Moreover, the collaboration within an interdisciplinary cross-functional team enables you to see different perspectives on the same problem, thus pushing you to work together towards a unified solution. The DPS coaches provide you with support and valuable insights throughout the journey. They have been an invaluable third eye for us, noticing aspects that we within the team couldn't at first. I highly recommend DPS to anyone interested in tackling real and interesting problem spaces by working together in a motivated team towards a common goal.” #GenAI #SupplyChain #Procurement ##LLM #RAG #DigitalProductSchool

🚀 𝐄𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞 𝐭𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞 𝐨𝐟 𝐀𝐧𝐢𝐦𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐓𝐞𝐧𝐜𝐞𝐧𝐭 𝐀𝐈 𝐋𝐚𝐛'𝐬 𝐓𝐨𝐨𝐧𝐂𝐫𝐚𝐟𝐭𝐞𝐫 Innovative Features: 1. 𝐓𝐫𝐚𝐧𝐬𝐜𝐞𝐧𝐝𝐢𝐧𝐠 𝐓𝐫𝐚𝐝𝐢𝐭𝐢𝐨𝐧𝐚𝐥 𝐈𝐧𝐭𝐞𝐫𝐩𝐨𝐥𝐚𝐭𝐢𝐨𝐧: ToonCrafter goes beyond conventional methods, tackling the challenges of cartoon animation with finesse. 2. 𝐓𝐫𝐚𝐧𝐬𝐜𝐞𝐧𝐝𝐢𝐧𝐠 𝐓𝐫𝐚𝐝𝐢𝐭𝐢𝐨𝐧𝐚𝐥 𝐈𝐧𝐭𝐞𝐫𝐩𝐨𝐥𝐚𝐭𝐢𝐨𝐧: Inspired by cutting-edge LDMs like DynamiCrafter, ToonCrafter seamlessly integrates live-action video motion priors into its generative framework. 3. 𝐑𝐞𝐬𝐨𝐥𝐯𝐢𝐧𝐠 𝐃𝐨𝐦𝐚𝐢𝐧 𝐆𝐚𝐩 𝐈𝐬𝐬𝐮𝐞𝐬: By adapting LDMs to the unique domain of cartoons, ToonCrafter ensures fidelity in every frame, overcoming traditional limitations. 4. 𝐏𝐫𝐞𝐬𝐞𝐫𝐯𝐢𝐧𝐠 𝐅𝐢𝐧𝐞 𝐃𝐞𝐭𝐚𝐢𝐥𝐬 𝐰𝐢𝐭𝐡 𝐚 𝟑𝐃 𝐃𝐞𝐜𝐨𝐝𝐞𝐫: Our dual-reference-based 3D decoder preserves intricate details lost in compressed latent spaces, ensuring stunning results. 5. 𝐄𝐦𝐩𝐨𝐰𝐞𝐫𝐢𝐧𝐠 𝐔𝐬𝐞𝐫 𝐂𝐨𝐧𝐭𝐫𝐨𝐥 𝐰𝐢𝐭𝐡 𝐚 𝐅𝐥𝐞𝐱𝐢𝐛𝐥𝐞 𝐒𝐤𝐞𝐭𝐜𝐡 𝐄𝐧𝐜𝐨𝐝𝐞𝐫: ToonCrafter's flexible sketch encoder puts interactive control over interpolation outcomes in the hands of creators, sparking creativity. Key Benefits: - 𝐕𝐢𝐬𝐮𝐚𝐥𝐥𝐲 𝐂𝐨𝐧𝐯𝐢𝐧𝐜𝐢𝐧𝐠 𝐚𝐧𝐝 𝐍𝐚𝐭𝐮𝐫𝐚𝐥 𝐃𝐲𝐧𝐚𝐦𝐢𝐜𝐬: ToonCrafter's innovative approach produces animations with unparalleled realism and fluidity. - 𝐄𝐟𝐟𝐞𝐜𝐭𝐢𝐯𝐞 𝐇𝐚𝐧𝐝𝐥𝐢𝐧𝐠 𝐨𝐟 𝐃𝐢𝐬-𝐨𝐜𝐜𝐥𝐮𝐬𝐢𝐨𝐧: From complex movements to intricate details, ToonCrafter excels in capturing the essence of every scene. - 𝐍𝐨𝐭𝐚𝐛𝐥𝐞 𝐒𝐮𝐩𝐞𝐫𝐢𝐨𝐫𝐢𝐭𝐲 𝐎𝐯𝐞𝐫 𝐂𝐨𝐦𝐩𝐞𝐭𝐢𝐭𝐨𝐫𝐬: Comparative evaluations confirm the remarkable performance of ToonCrafter against existing methods. Join us in embracing the future of AI-driven animation. The revolution has begun! 🌟 Follow AI Makerspace by Digital Product School for the latest updates and insights into AI technology! 🚀🤖 #AI #innovation #technology #animation #ToonCrafter

🎬 Unveiling the Next Generation of AI Animation! 🌟 Microsoft Research has just unveiled their latest groundbreaking project: 𝑽𝑨𝑺𝑨-1. Here's why you should be excited: • 𝐋𝐢𝐟𝐞𝐥𝐢𝐤𝐞 𝐅𝐚𝐜𝐢𝐚𝐥 𝐀𝐧𝐢𝐦𝐚𝐭𝐢𝐨𝐧: VASA-1 generates incredibly realistic talking faces from static images and audio clips. Imagine avatars that not only speak but also exhibit natural facial expressions and head movements. • 𝐇𝐨𝐥𝐢𝐬𝐭𝐢𝐜 𝐃𝐲𝐧𝐚𝐦𝐢𝐜𝐬: This AI captures a wide range of facial nuances, ensuring that the avatars are not just visually appealing but also authentically lifelike. • 𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧: Capable of producing 512x512 resolution videos at up to 40 frames per second with negligible latency, making it perfect for real-time applications. • 𝐀𝐝𝐯𝐚𝐧𝐜𝐞𝐝 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠: Utilizing a Diffusion Transformer model, VASA-1 is trained on massive datasets to deliver nuanced facial dynamics across different identities. • 𝐂𝐮𝐬𝐭𝐨𝐦𝐢𝐳𝐚𝐛𝐥𝐞 𝐀𝐯𝐚𝐭𝐚𝐫𝐬: Offers optional control signals such as eye gaze direction, head-to-camera distance, and emotional expressions for tailored interactions. • 𝐄𝐭𝐡𝐢𝐜𝐚𝐥 𝐂𝐨𝐧𝐬𝐢𝐝𝐞𝐫𝐚𝐭𝐢𝐨𝐧𝐬: Acknowledging the potential for misuse, Microsoft ensures that VASA-1 is rolled out responsibly. Join us in embracing the future of AI. The revolution has begun, and VASA-1 is leading the charge. 🌟 Follow AI Makerspace by Digital Product School for the latest updates and insights into AI technology! 🚀🤖 #AI #MachineLearning #VASA1 #MicrosoftResearch #Innovation #DigitalTransformation #AIRevolution #FutureTech #TechInnovation

🚀 The Future of AI is Here! 🌟 OpenAI recently unveiled their latest breakthrough model: 𝑮𝑷𝑻-4𝒐. Here's why you should be buzzing about it: • 𝐎𝐦𝐧𝐢-𝐌𝐨𝐝𝐚𝐥 𝐌𝐚𝐬𝐭𝐞𝐫𝐲: Handles text, speech, and video effortlessly. The "o" in GPT-4o stands for “omni” 🌐. • 𝐄𝐧𝐡𝐚𝐧𝐜𝐞𝐝 𝐈𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐨𝐧: Engage with AI more naturally across voice, text, and vision. Think of a future where machines understand and respond to you like never before. • 𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐢𝐯𝐞𝐧𝐞𝐬𝐬: Interrupt and interact with GPT-4o seamlessly, even picking up on voice nuances for emotive responses, including singing! 🎤 • 𝐕𝐢𝐬𝐮𝐚𝐥 𝐆𝐞𝐧𝐢𝐮𝐬: Analyze and respond to images instantly. From identifying brands to decoding software screens, GPT-4o has you covered. • 𝐌𝐮𝐥𝐭𝐢𝐥𝐢𝐧𝐠𝐮𝐚𝐥 𝐌𝐚𝐫𝐯𝐞𝐥: Supports over 50 languages with enhanced performance 🌍. • 𝐒𝐩𝐞𝐞𝐝 & 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐜𝐲: Twice as fast and half the price of GPT-4 Turbo, making cutting-edge AI more accessible 🚀. OpenAI has also refreshed the ChatGPT UI, introducing a more conversational home screen and message layout. Plus users get early access to these new tools, along with the desktop app for macOS that allows for quick interactions and screenshot discussions. The GPT-4o API with voice support is launching first to a select group, ensuring a safe and responsible rollout. Join us in embracing the future of AI. The revolution has begun, and GPT-4o is leading the charge. 🌟 Follow AI Makerspace by Digital Product School for the latest updates and insights into AI technology! 🚀🤖 #gpt #AI_Makerspace #chatgpt #ai

🚀 #SuccessStory: Team WhatSAP at Digital Product School by UnternehmerTUM and in collaboration with SAP, developed X-Factor, to pioneer a new generation of sales forecasting. This product uses #GenAI to extract insights from social media and seamlessly integrates them into predictive algorithms. 🤝 User Problem Sales forecasts are impacted by diverse factors such as economic trends, social changes, supply disruptions, competition, and regulations. However, AI engineers face the challenge of collecting and organizing dispersed online data to enhance the accuracy of forecasts. 💡 Solution Idea The Solution enables users to: 👉 Select a specific industry and identify factors influencing sales. 👉 Extract daily information from the web to create an insight dataset. 👉 Use the structured data to recognize patterns and make precise predictions. 👉 Use the structured data to fine-tune and adapt sales forecasting models. 🎯 Applications of AI 👉 Interactions to generate user queries. 👉 Search and retrieve information from images, diagrams, and tables. 👉 Customize data extraction based on user feedback and queried sales impact. 👉 Generate sample codes for predictive procedures. 👩💻 Tools Used #LLM, #typescript, #react, #AWSlambda, #Figma 🤗 WhatSAP Team Members Harshinee Baaskar: Product Manager Jacker She: Interaction Designer Tanisha Sharma: AI Engineer Robert M.: Software Engineer Mehmet Ali Altunsoy: Software Engineer Wenjun X. Xue: Software Engineer 💎 Testimony by Tanisha Sharma "DPS has been an incredibly fulfilling experience for me. Beyond the innovative products and cutting-edge technology, what truly sets DPS apart is its commitment to fostering personal and professional growth. From day one, I felt supported by a team that values collaboration, creativity, and continuous learning. Whether you're a seasoned professional or just starting your career, DPS provides a dynamic environment where your ideas are not only welcomed but encouraged. Here, you're not just a participant, you're part of a passionate set of people dedicated to pushing boundaries and making a real impact. If you're looking for more than just a program, but a place where you can thrive and make a difference, I highly encourage you to join DPS. 💖 " Join Digital Product School by UnternehmerTUM and be part of the AI innovation journey. Apply from our website: https://lnkd.in/e8QShpDN Follow AI Makerspace by Digital Product School for more insights into our world of human-centred AI entrepreneurship.