Nyun AI

We're #hiring a new Deep Learning Engineer Intern in Karnataka. Apply today or share this post with your network.

🚀 Join Us as a Deep Learning Developer Intern! 🚀 We're excited to offer internship opportunities for our open-source project, #Nyuntam, on GitHub! Nyuntam is dedicated to compressing large models like LLMs to enable foundational AI on edge devices. See here: https://lnkd.in/gfSuhJGQ We're looking for passionate Deep Learning Developer Interns to contribute to Nyuntam by adding new features under the guidance of our expert team. Why Join? Roles: Full-time or part-time Perks: Co-author a paper for a top ML conference upon successful completion and integration of assigned tasks. Stipend: Generous compensation If you're ready to work on cutting-edge AI optimization and gain recognition in the community, this opportunity is for you! 📋 Apply here: https://lnkd.in/dZ4bn_eX Be part of the journey to bring AI to the edge! ✨ #AI #DeepLearning #Internships #OpenSource #AIOnEdge #MachineLearning

As Generative AI expands into edge computing—enabling real-time decision-making on smartphones, IoT devices, and more—inaccuracy has become a critical concern. According to McKinsey's 2024 report, many businesses adopting GenAI on the edge struggle with inaccurate outputs, especially in tasks such as summarization and content generation. These inaccuracies can have severe consequences in sectors like healthcare or customer service, where reliability is paramount.This challenge is further exacerbated by the hardware limitations of edge devices, which make it difficult to run large, complex models without performance trade-offs. Many organizations are working to mitigate this by fine-tuning models for specific tasks and investing in better governance and explainability to ensure trust in AI-generated results. At Nyun AI, we are focused on deploying highly efficient GenAI models on the edge while maintaining the integrity and performance of the models. Our edge-optimized solutions prioritize accuracy and speed, ensuring that businesses can harness the power of GenAI without compromising on quality. If you're exploring edge AI solutions, let's connect and discuss how we can drive impactful results together. Ping us at connect@nyunai.com. Link: https://lnkd.in/eADN79Cd

𝗧𝗵𝗮𝗻𝗸 𝗬𝗼𝘂 𝗳𝗼𝗿 𝗝𝗼𝗶𝗻𝗶𝗻𝗴 𝗢𝘂𝗿 𝗪𝗲𝗯𝗶𝗻𝗮𝗿! We were overwhelmed by the response to our recent webinar! As requested, you can now access the 𝗣𝗣𝗧 𝘀𝗹𝗶𝗱𝗲𝘀 from the session below. It was great connecting with all of you and sharing insights. During the event, we also had the opportunity to showcase our open-source tool, 𝗡𝘆𝘂𝗻𝘁𝗮𝗺. For those who missed it or want to explore further, check out the repo : https://lnkd.in/gfSuhJGQ. But that’s not all! Stay tuned and 𝗳𝗼𝗹𝗹𝗼𝘄 𝗼𝘂𝗿 𝗽𝗮𝗴𝗲 for updates – we’ll be announcing our 𝗻𝗲𝘅𝘁 𝘄𝗲𝗯𝗶𝗻𝗮𝗿 𝘃𝗲𝗿𝘆 𝘀𝗼𝗼𝗻 with even more exciting demos and in-depth discussions! Don’t miss out. #𝗪𝗲𝗯𝗶𝗻𝗮𝗿 #𝗡𝘆𝘂𝗻𝘁𝗮𝗺 #𝗢𝗽𝗲𝗻𝗦𝗼𝘂𝗿𝗰𝗲 #𝗧𝗲𝗰𝗵𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆 #𝗡𝗲𝘅𝘁𝗪𝗲𝗯𝗶𝗻𝗮𝗿 #𝗖𝗼𝗺𝗽𝗮𝗻𝘆𝗨𝗽𝗱𝗮𝘁𝗲𝘀 #𝗙𝗼𝗹𝗹𝗼𝘄𝗨𝘀

We are happy to share that we wrapped up NyunAI's first webinar in collaboration with 𝗧𝗿𝗮𝗻𝘀𝗺𝘂𝘁𝗲 𝗔𝗜! The session, led by Shubham Kushwaha, was a deep dive into the fascinating world of creating 𝘂𝗹𝘁𝗿𝗮-𝗰𝗼𝗺𝗽𝗮𝗰𝘁 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 (𝗟𝗟𝗠𝘀) using advanced model quantization techniques. 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀 𝗳𝗿𝗼𝗺 𝘁𝗵𝗲 𝗪𝗲𝗯𝗶𝗻𝗮𝗿: • Shubham Kushwaha simplified the complex process of making LLMs more efficient and compact without compromising performance. • A special focus was on 𝗡𝘆𝘂𝗻𝗔𝗜’𝘀 𝗹𝗮𝘁𝗲𝘀𝘁 𝘁𝗼𝗼𝗹, 𝗡𝘆𝘂𝗻𝘁𝗮𝗺, designed for easy and effective model compression, making the development of ultra-compact LLMs accessible for everyone. We’ve got some great insights from the event that we can’t wait to share! Check out the attached screenshots from the session. If you have any questions, doubts, or would like to discuss this topic further, feel free to reach out to our CTO Arnav Chavan or the presenter Shubham Kushwaha 𝗧𝗵𝗶𝘀 𝗶𝘀 𝗷𝘂𝘀𝘁 𝘁𝗵𝗲 𝗯𝗲𝗴𝗶𝗻𝗻𝗶𝗻𝗴! Stay tuned and 𝗸𝗲𝗲𝗽 𝗳𝗼𝗹𝗹𝗼𝘄𝗶𝗻𝗴 𝗡𝘆𝘂𝗻𝗔𝗜 as we continue to bring you more innovative sessions around 𝗺𝗼𝗱𝗲𝗹 𝗰𝗼𝗺𝗽𝗿𝗲𝘀𝘀𝗶𝗼𝗻, 𝗟𝗟𝗠𝘀, and beyond. #AI #LLMs #ModelCompression #NyunAI #Quantization #Nyuntam #Innovation #TechWebinar #TransmuteAI #DeepLearning

🚨 𝟭 𝗺𝗼𝗿𝗲 𝗱𝗮𝘆 𝘁𝗼 𝗴𝗼! 🚨 Don’t miss out on exploring 𝘂𝗹𝘁𝗿𝗮-𝗰𝗼𝗺𝗽𝗮𝗰𝘁 𝗟𝗟𝗠 𝘀𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗲𝘀 𝘄𝗶𝘁𝗵 𝘂𝗽 𝘁𝗼 𝟴𝘅 𝗺𝗲𝗺𝗼𝗿𝘆 𝗿𝗲𝗱𝘂𝗰𝘁𝗶𝗼𝗻. Join us tomorrow for a deep dive into advanced quantization algorithms and unlock faster, more efficient model deployment! 𝗗𝗮𝘁𝗲 & 𝗧𝗶𝗺𝗲: 12th September 2024, 9:00 PM IST | 8:30 AM PT 𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿 𝗡𝗼𝘄: https://shorturl.at/RKwGh #𝗔𝗜 #𝗠𝗮𝗰𝗵𝗶𝗻𝗲𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 #𝗠𝗼𝗱𝗲𝗹𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 #𝗤𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 #𝗟𝗟𝗠𝘀 #𝗧𝗲𝗰𝗵𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 #𝗪𝗲𝗯𝗶𝗻𝗮𝗿

🗓 Just a gentle reminder! Our webinar on 𝗮𝗰𝗵𝗶𝗲𝘃𝗶𝗻𝗴 𝘂𝗹𝘁𝗿𝗮-𝗰𝗼𝗺𝗽𝗮𝗰𝘁 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 (𝗟𝗟𝗠𝘀) is just 3 days away! 𝗗𝗮𝘁𝗲: 12th September 2024, 9:00 PM IST | 8:30 AM PT 𝗗𝗼𝗻’𝘁 𝗺𝗶𝘀𝘀 𝗼𝘂𝘁. 𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿 𝗵𝗲𝗿𝗲: https://shorturl.at/RKwGh You can also explore Nyuntam before the event: https://lnkd.in/gfSuhJGQ #𝗠𝗮𝗰𝗵𝗶𝗻𝗲𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 #𝗔𝗜 #𝗠𝗼𝗱𝗲𝗹𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 #𝗟𝗟𝗠𝘀 #𝗧𝗲𝗰𝗵𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 #𝗪𝗲𝗯𝗶𝗻𝗮𝗿

We just published a new blog on 𝗠𝗲𝗱𝗶𝘂𝗺 about how to optimize large language models (LLMs) like 𝗟𝗹𝗮𝗺𝗮𝟯.𝟭-𝟴𝗕 using the 𝗪𝟰𝗔𝟴𝗞𝗩𝟰 𝗾𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 technique from LMQuant. In the blog, we walk through how this approach can help reduce model size, improve throughput, and cut down on computational costs, especially for large-scale deployments! 𝗞𝗲𝘆 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀: • How to apply 𝟰-𝗯𝗶𝘁 𝘄𝗲𝗶𝗴𝗵𝘁𝘀 (𝗪𝟰), 𝟴-𝗯𝗶𝘁 𝗮𝗰𝘁𝗶𝘃𝗮𝘁𝗶𝗼𝗻𝘀 (𝗔𝟴), and 𝟰-𝗯𝗶𝘁 𝗸𝗲𝘆-𝘃𝗮𝗹𝘂𝗲 𝗰𝗮𝗰𝗵𝗲𝘀 (𝗞𝗩𝟰). • Achieving up to 𝟯.𝟱𝘅 𝘀𝗽𝗲𝗲𝗱 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝗺𝗲𝗻𝘁𝘀 on GPUs like the A100 and L40S with the 𝗤𝗦𝗲𝗿𝘃𝗲 system. • Step-by-step guide to model quantization, installation, and performance evaluation. 𝗟𝗶𝗻𝗸 𝘁𝗼 𝘁𝗵𝗲 𝗯𝗹𝗼𝗴 𝗽𝗼𝘀𝘁: https://bit.ly/4dIWbJf 𝗟𝗶𝗻𝗸 𝘁𝗼 𝘁𝗵𝗲 𝗽𝗮𝗽𝗲𝗿: https://bit.ly/4g787WK 𝗟𝗶𝗻𝗸 𝘁𝗼 𝗡𝘆𝘂𝗻𝘁𝗮𝗺 𝗥𝗲𝗽𝗼𝘀𝗶𝘁𝗼𝗿𝘆: https://bit.ly/4ggbISl Check out the blog to learn how you can maximize performance while minimizing computational overhead, and explore Nyuntam to start optimizing your models today! #𝗔𝗜 #𝗠𝗮𝗰𝗵𝗶𝗻𝗲𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 #𝗠𝗼𝗱𝗲𝗹𝗖𝗼𝗺𝗽𝗿𝗲𝘀𝘀𝗶𝗼𝗻 #𝗡𝘆𝘂𝗻𝘁𝗮𝗺 #𝗢𝗽𝗲𝗻𝗦𝗼𝘂𝗿𝗰𝗲

🚀 𝗝𝗼𝗶𝗻 𝗢𝘂𝗿 𝗘𝘅𝗰𝗹𝘂𝘀𝗶𝘃𝗲 𝗪𝗲𝗯𝗶𝗻𝗮𝗿 𝗼𝗻 𝗔𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗤𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗧𝗲𝗰𝗵𝗻𝗶𝗾𝘂𝗲𝘀 𝘄𝗶𝘁𝗵 𝗡𝘆𝘂𝗻𝘁𝗮𝗺! 🚀 Are you ready to dive deep into the world of AI model compression? We’re excited to invite you to our upcoming webinar where we’ll explore complex quantization schemes, including the innovative AQLM algorithm combined with the powerful PV Tuning for 2-bit task-aware LLMs. This session is perfect for engineers and researchers looking to enhance their understanding and practical skills in model optimization. 𝗪𝗵𝗮𝘁 𝗬𝗼𝘂 𝗪𝗶𝗹𝗹 𝗟𝗲𝗮𝗿𝗻: • In-depth understanding of AQLM, PV Tuning and such advanced quantization algorithms and its impact on model efficiency. • Step-by-step guide on implementing them using our open-source library, Nyuntam, to bring cutting-edge compression to your projects. 𝗪𝗲𝗯𝗶𝗻𝗮𝗿 𝗗𝗲𝘁𝗮𝗶𝗹𝘀: • 📅 Date: 12th September 2024 • 🕒 Time: 9:00 PM IST, 8:30 AM PT • 🌐 Location: Online Don’t miss this opportunity to boost your models’ performance while significantly reducing computational costs. Whether you’re a seasoned pro or just getting started, this webinar will equip you with the knowledge to implement advanced quantization techniques effectively. 👉 𝗥𝗲𝘀𝗲𝗿𝘃𝗲 𝘆𝗼𝘂𝗿 𝘀𝗽𝗼𝘁 𝗻𝗼𝘄! https://bit.ly/3Moxju6 📣 Feel free to share this with your network and anyone who might benefit from this advanced learning session! Looking forward to seeing you there!

𝗡𝘆𝘂𝗻𝘁𝗮𝗺, an open-source toolkit designed by Nyun can help you compress and adapt AI models with ease! Whether you're working with large language models like Llama3-8B or any other AI model, Nyuntam offers practical tools that streamline your workflow, making powerful models more efficient and deployable. In our latest blog, we dive deep into how Nyuntam’s 𝗔𝗱𝗱𝗶𝘁𝗶𝘃𝗲 𝗤𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗟𝗮𝗿𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 (𝗔𝗤𝗟𝗠) + 𝗣𝗩 𝗧𝘂𝗻𝗶𝗻𝗴 technique can compress models like Llama3-8B down to 2-bit weights without sacrificing downstream task performance, making them perfect for resource-constrained environments. Link to the blog post: https://bit.ly/3MuRFlt Check out the blog to learn how you can maximize performance while minimizing computational overhead, and explore Nyuntam to start optimizing your models today! #AI #MachineLearning #ModelCompression #Nyuntam #OpenSource