Mathos AI

How to prove that I didn't pay for this 🆘? JKJK 🤭🤭🤭 Feel super grateful! Also, I would like to share below the story I told Forbes Journalist Katherine Love when I was reached out as the nominee a few months ago: # Brief Background I have loved math since I was young. I was admitted to a top university in China at just 14 years old due to my strong performance in math competitions. Then I moved to the U.S. in 2018 to further study Natural Language Processing (NLP) at Carnegie Mellon University, where I experienced firsthand the transformative power of language models. # Motivation As ChatGPT completely transformed the way people acquire information, I recognized an unprecedented opportunity to revolutionize education. Meanwhile, I noticed many people are still suffering from learning math, while I have loved and benefited a lot from math. So I decided to build something that can help. # What We Built Our featured app is a multimodal personalized AI math tutor that allows students to Circle & Ask, Draw, Scratch, and Talk, all at once, seamlessly, and learn at their optimized pace. It’s like having a supportive math teacher right by your side with infinite patience. We have also developed a comprehensive RAG system to answer questions in a way that is aligned with the curriculum and offer that as an API to empower next-generation personalized learning platforms. # Relevant Facts and Figures We have crossed 1M users within a year by organic growth. We have raised 1.84M (oversubscribed!) very quickly. Our model has 15% accuracy improvement over GPT4o on high school and college level math problems. We have also partnered with some EdTech companies (including 𝗜𝗣𝗢 𝗰𝗼𝗺𝗽𝗮𝗻𝗶𝗲𝘀) to offer API and personalized learning solutions. # Mission We want to empower everyone to learn anything from anywhere. Starting with the AI Math Tutor, we are envisioning a future where quality, affordable, and personalized learning experiences are accessible to all people. Looking forward, we will keep expanding and utilize @Mathos to empower teachers, school districts, and parents to best understand and assist their kids, in order to improve personalized education at population level.

Extremely honored to be recognized as #ForbesUnder30, North America, Class of 2025. It's the best timing for LLMs, as well as a unprecedented opportunity for future education. Stay tuned!

Paper accepted to @COLM_conf! If you don't know CoLM, it's Neurips-level, but the 𝘧𝘪𝘳𝘴𝘵 𝘦𝘷𝘦𝘳 𝘤𝘰𝘯𝘧𝘦𝘳𝘦𝘯𝘤𝘦 𝘧𝘰𝘳 𝘓𝘓𝘔𝘴. Wild. Publishing papers while building a 𝘀𝘁𝗮𝗿𝘁𝘂𝗽 with 𝟭𝗠+ users is never easy. We developed 𝗦𝘁𝗮𝘁𝗲𝗙𝗹𝗼𝘄, a framework to better control LLM to solve highly complex tasks, with significant performance gain but 5x cheaper cost. It also allows further improvement by combining with iterative refining methods (e.g. Reflexion). 𝗦𝘁𝗮𝘁𝗲𝗙𝗹𝗼𝘄 was also quickly integrated with @pyautogen and "noticeably improved" (cited from a user testimonial) the performance (28.3k ⭐). AutoGen: github.com/microsoft/auto… StateFlow: github.com/yiranwu0/State… Thanks my amazing co-authors especially @YiranWu18 who iterates this idea over and over again with me (well, we also iterated the paper name a lot 🤭), and @ShaokunZhang1 @qingyun_wu @Chi_Wang_❤️, and my friends who offer meaningful discussions @hanjundai @ericzelikman @jyangballin @cindy_x_wu Here are more insights:

Join us for 3rd lecture on agent frameworks in our LLM Agents MOOC, @chi_wang_ @GoogleDeepMind & @jerryjliu0 @llama_index, 3:10pm Sep 23: llmagents-learning.org, with 7K+ registered learners, 3.5K+ members in discord! 🎉📢 Huge thanks to @ShunyuYao12 @OpenAI for 2nd lecture on LLM agents overview: youtube.com/watch?v=RM6ZAr…