2026

AI-Based System for Facilitating Awareness in Everyday Environments
AI-Based System for Facilitating Awareness in Everyday Environments

Shengxin Li

Lead Researcher & Full-Stack Developer · Master's Research Project · 2026 - Present

My research project proposes a multilingual web system that helps people reflect on overlooked details along familiar routes. I designed and built a staged experience from unaided recall and video-aided reseeing to AI-supported discovery and revisable co-creation using Next.js and TypeScript. Its multimodal pipeline combines OpenAI frame-level vision analysis with Gemini full-video reasoning to surface environmental and seasonal observations grounded in video evidence. I also implemented participant-isolated authentication, data storage, and interaction logging with Supabase and private Cloudflare R2 storage, and independently led a two-week longitudinal field study with 15 participants.

[Demo] [Slides (progress ongoing)]

AI-Based System for Facilitating Awareness in Everyday Environments
AI-Based System for Facilitating Awareness in Everyday Environments

Shengxin Li

Lead Researcher & Full-Stack Developer · Master's Research Project · 2026 - Present

My research project proposes a multilingual web system that helps people reflect on overlooked details along familiar routes. I designed and built a staged experience from unaided recall and video-aided reseeing to AI-supported discovery and revisable co-creation using Next.js and TypeScript. Its multimodal pipeline combines OpenAI frame-level vision analysis with Gemini full-video reasoning to surface environmental and seasonal observations grounded in video evidence. I also implemented participant-isolated authentication, data storage, and interaction logging with Supabase and private Cloudflare R2 storage, and independently led a two-week longitudinal field study with 15 participants.

[Demo] [Slides (progress ongoing)]

Go Touch Grass: AI-Assisted Low-Barrier Bilingual Community Volunteer Platform
Go Touch Grass: AI-Assisted Low-Barrier Bilingual Community Volunteer Platform

Shengxin Li*, Claudio González Alanís*, Zixi Zhu*, Wenkai Jia* (* equal contribution)

Full-Stack Developer & Project Manager · Practice Information Systems · Spring 2026

Go Touch Grass connects volunteers and community organizers through a Japanese and English platform. I built the end-to-end volunteer experience with Next.js and TypeScript, including role-based registration, opportunity discovery, task-level applications, status tracking, and pre-event communication. For organizers, I developed a bilingual guided form with Gemini-assisted drafting, using Prisma and PostgreSQL to store multi-task event listings and application states. As project manager, I led sprint planning, user-story authoring, and backlog management while coordinating reviews and contributing to debugging and testing across features.

[Slides]

Go Touch Grass: AI-Assisted Low-Barrier Bilingual Community Volunteer Platform
Go Touch Grass: AI-Assisted Low-Barrier Bilingual Community Volunteer Platform

Shengxin Li*, Claudio González Alanís*, Zixi Zhu*, Wenkai Jia* (* equal contribution)

Full-Stack Developer & Project Manager · Practice Information Systems · Spring 2026

Go Touch Grass connects volunteers and community organizers through a Japanese and English platform. I built the end-to-end volunteer experience with Next.js and TypeScript, including role-based registration, opportunity discovery, task-level applications, status tracking, and pre-event communication. For organizers, I developed a bilingual guided form with Gemini-assisted drafting, using Prisma and PostgreSQL to store multi-task event listings and application states. As project manager, I led sprint planning, user-story authoring, and backlog management while coordinating reviews and contributing to debugging and testing across features.

[Slides]

VR Japanese Daily-Life Training: Situated Japanese Practice in VRChat
VR Japanese Daily-Life Training: Situated Japanese Practice in VRChat

Shengxin Li*, Yiyan Deng*, Siyi Wang*, Zixi Zhu*, Wenkai Jia* (* equal contribution)

Developer & Project Manager · Practice Information Systems · Spring 2026

This VRChat convenience-store prototype connects Japanese conversation with an everyday shopping procedure. I implemented proximity-triggered interfaces, voice- or click-based choices, and synchronized Guided/Free practice states using Unity and UdonSharp. A Python/FastAPI pipeline with faster-Whisper advances guided dialogue from live speech, using step-specific prompts, confidence checks, duplicate suppression, and TTS-echo filtering. For open-ended practice, I integrated OpenAI structured outputs and speech synthesis to separate concise language feedback from the clerk's next reply and route Japanese speech through VRChat's microphone controls for repeated role-play.

[Slides]

VR Japanese Daily-Life Training: Situated Japanese Practice in VRChat
VR Japanese Daily-Life Training: Situated Japanese Practice in VRChat

Shengxin Li*, Yiyan Deng*, Siyi Wang*, Zixi Zhu*, Wenkai Jia* (* equal contribution)

Developer & Project Manager · Practice Information Systems · Spring 2026

This VRChat convenience-store prototype connects Japanese conversation with an everyday shopping procedure. I implemented proximity-triggered interfaces, voice- or click-based choices, and synchronized Guided/Free practice states using Unity and UdonSharp. A Python/FastAPI pipeline with faster-Whisper advances guided dialogue from live speech, using step-specific prompts, confidence checks, duplicate suppression, and TTS-echo filtering. For open-ended practice, I integrated OpenAI structured outputs and speech synthesis to separate concise language feedback from the clerk's next reply and route Japanese speech through VRChat's microphone controls for repeated role-play.

[Slides]