Product

In the rapidly evolving landscape of artificial intelligence, we’re witnessing a fundamental shift in how we consume and interact with knowledge. While early AI applications focused primarily on content summarization and modal conversion, the next generation of AI-native products promises something far more transformative: the ability to create truly personalized learning experiences that adapt to individual needs, interests, and cognitive patterns. The Limitations of Current Approaches Today’s AI content tools largely excel at taking existing information and reformatting it into different modalities. We can convert text to audio, create video summaries, or generate podcast-style conversations from written material. However, as Large Language Model context windows continue to expand, simple content summarization becomes increasingly commoditized. The real value lies not in these mechanical transformations, but in the creative synthesis and novel perspectives that emerge when AI systems understand both the content and the consumer. ...

This week, I wanted to organize my thoughts about AI-generated content (AIGC) and creativity-related products from the past few months. Rather than focusing solely on my own projects, I’d like to explore the foundational aspects of AI product design, interspersing examples from my recent work. First, I want to emphasize that technology is merely a tool intended to better serve business needs. If it doesn’t significantly improve efficiency, traditional methods may be more appropriate. Second, despite the many imaginative possibilities of current technology, applications should ultimately be guided by user needs. Finally, AI technologies and markets evolve rapidly, making predictions difficult to validate, but exploring content understanding and generation remains an intriguing challenge. ...

Building Intelligent Content Discovery: A Multi-Modal Recommendation System for Podcast Consumption Modern content platforms face a fundamental challenge: how to help users discover relevant, high-quality content that matches their interests while avoiding the trap of information overload. This challenge becomes particularly acute in podcast consumption, where users need to find content that not only aligns with their interests but also fits different consumption contexts and scenarios. An ideal recommendation system addresses these core problems through a sophisticated approach that combines content-based filtering, collaborative filtering, and contextual understanding to create a truly personalized discovery experience. ...

Leverage generative AI capabilities for creative script ideation and video ad creation. (Worked on agentic workflows and interface optimization) https://ads.tiktok.com/business/copilot/standalone?locale=en&deviceType=pc Credits: TikTok Creative Team Building Agentic Workflows From LLMs to Agents The transition from LLMs to Agents has become a consensus in the AI community, representing an improvement in complex task execution capabilities. However, helping users fully utilize Agent capabilities to achieve tenfold efficiency gains requires careful workflow design. These workflows aren’t merely a presentation of parallel capabilities, but seamless integrations with human-in-the-loop quality assurance. This document uses Typeface as a reference to explain why a clear primary workflow is necessary, as well as design approaches for functional extensions. ...

Analyze thousands of tiktoks to provide actionable trends & insights for key agencies. (Worked on multi-modal content understanding) To be released on TikTok Creative Center (https://ads.tiktok.com/business/creativecenter/pc/en) Credits: TikTok Creative Team TikTok Insight Spotlight Launches at Advertiser Summit Jun 3 - Excited to share that TikTok Insight Spotlight, a product I worked on during my time at TikTok, was officially unveiled at the company’s annual advertiser summit on June 3rd, 2025. The Verge covered the launch extensively, highlighting the AI-driven capabilities I helped develop. ...

Marrrket is an AI-powered second-hand marketplace platform targeting North American university students, initially focusing on Chinese international students. The platform aims to solve the inefficiency in the current second-hand market by simplifying the listing process through AI-generated product descriptions from images and minimal user input. By reducing friction in the listing process, Marrrket will increase the overall volume of second-hand items available in the market, creating a more efficient marketplace for both buyers and sellers. The platform’s innovation centers on using artificial intelligence to dramatically lower the barrier to entry for sellers, which is hypothesized to be the primary constraint on market growth. ...

Multi-agent system for cross boarder e-commerce sales automation. Co-founder and head of product. https://cognogpt.com Cogno+ is dedicated to revolutionizing global e-commerce by empowering brands with AI-driven assistance that offers seamless, personalized customer experiences. Our mission is to serve as the digital bridge between brands and customers, enhancing interactions and transactions across international markets, and allowing the brand to increase conversion and upsell while reducing time and money spent on manual customer service. ...

AI Homework helper with advanced reasoning and visualization for all school subjects. (Worked on LLM Reasoning) https://www.gauthmath.com/ Credits: Lexi Ling, Gauth Team

Extracting meaningful insights from unstructured multi-modal content Most times, the challenge isn’t collecting information but extracting value from it. By developing systems that can analyze unstructured multi-modal content (text, images, video, audio), we can extract actionable insights.

Inspiring insights, amplifying voices. (crowdlistening.com) From Content Aggregation to Original Research Crowdlistening transforms large-scale social conversations into actionable insight by integrating llm reasoning with extensive model context protocol(MCP) capabilities. While being able to quantatively analyze large volumes of data is already an interesting task, our focus is not just on content analysis at scale, but rather conducting original research directly from raw social data, generating insights that haven’t yet appeared in established reporting. ...