Serve community and humanity by applying technology.
Explore more about me
Adventure X
I led a non-profit hackathon with over 2,000 visitors, 220 participants, and 82 companies. As co-organizer, I managed a diverse team, overseeing logistics, sponsorships, and event programming. My team delivered 20+ workshops, offering hands-on experiences, and facilitated 60+ innovative projects, with 10+ receiving potential investment term sheets.
Shaman Shendiao Performance
Shendiao is a special type of folk song in Northeastern Shaman culture in China. As one of the few Northeasterners at school, I led my group in learning songs and culture from the Northeastern shaman and performing at events. I analyzed regional myths, exploring their evolution and cultural values. I also cook Northeastern dishes, attracting my peers to visit my hometown.
Friendsville Children Center
This NGO supports autistic children with tailored education and psychological counseling, overcoming barriers to specialized resources. Founded to promote inclusion, it fosters personal growth, social integration, and community awareness.
United AI Masters
I led an initiative promoting ethical AI and inclusivity, collaborating with students from six schools to recruit 380 members and gather over 150 signatures for the AI Honor Code. I organized efforts, delegated tasks, and acted as the primary contact for participants. Currently, I'm working with Suno to create music to help children with ASD focus.
IELTS Writing Grader Model
I developed an AI-powered IELTS writing grader using Meta's Llama 3 model, fine-tuned on Tsinghua University's servers and Google Colab. With help from Tsinghua teachers, I successfully solved problems in server computing power and language encoding. I also created a dataset using IELTS textbooks for scoring essays.
Continuous Approximations for Improving Quantization Aware Training of LLMs
Tutored by Professor Guojie Song at Peking University, my team researched Quantization-Aware Training and the Straight Through Estimator to optimize neural network performance. Using three 4090 GPUs to train the model, my team achieved a 2.76% improvement on BoolQ and 5.47% on MMLU. The paper was submitted to the NeurIPS high school track.
Cornell University Pre-college
I studied "Social Media, Privacy, and Data Ethics" under Professor Faller, engaging in discussions on the interplay between technology and societal norms. I covered topics like privacy violations on social media, deepfake prevention, and bias in AI. I wrote a paper on privacy issues in datasets for training AI models, earning an A and 3 credits.
Internship in Mathos AI
I focused on promoting the product domestically and driving user acquisition. I conducted product evaluations and collected feedback from local teachers and institutions to guide improvements. I also represented the company at promotional events and tech showcases, aligning the product with market needs for AI-driven math education solutions.