Team
Kiwi
Project Concept
No description has been added yet.
Entry
Status: Submitted
Last saved: December 11 at 9:26 PM +08
Team Roster
Message board not available for this team yet.
Li Lin Team Lead RSVP Approved
Student at National University of Singapore
built the full Next.js Web UI, including prompt submission, progress feedback, and video delivery pages. They integrated Clerk for secure authentication and user-specific sessions, ensuring each generation is isolated per user. They connected the frontend to backend agent endpoints, handled API workflow, and designed the user interface flow for a smooth, intuitive experience. They ensured the system could run fully in the browser with no setup required from judges.
An AI engineer and product builder who loves creating intelligent agent systems that can reason, collaborate, and act autonomously. I work extensively with LLM-driven multi-agent frameworks like LangChain and LangGraph, and I enjoy building end-to-end systems using Python, FastAPI, TimescaleDB, Docker, and modern frontends.
I’ve developed multi-agent trading research platforms, personal AI memory/insight assistants, and behavior-aware productivity tools that integrate signals like Apple Watch data. My focus is always on turning complex AI capabilities into practical, user-centered applications.
I’m excited about pushing AI agents beyond simple chat — toward systems that think strategically, coordinate like expert teams, and deliver meaningful real-world impact.
I’m interested in building AI agents that enhance daily life—improving productivity, learning, wellbeing, and decision-making. I hope to connect with collaborators who want to create practical, human-centered AI products that help people focus better, make smarter choices, and live more intentionally.
I’m currently building MarketLens / Alpha-Trade, a multi-agent financial intelligence platform that combines LSTM price prediction, technical indicators, and autonomous research agents for real-time market reasoning. I’m also experimenting with a Cursor–Codex agent loop, enabling two models to collaborate, critique code, and generate iterative improvements directly inside the developer workflow.
YANG MIAO RSVP Approved
master at national university of singapore
integrated ElevenLabs for voice synthesis and implemented the MoviePy pipeline to merge narration with Veo-generated video scenes. They handled edge-case logic, stability improvements, and audio/video synchronization. They conducted end-to-end testing across diverse prompts to ensure consistent output quality. They also set up CodeRabbit for automated pull-request review, improving code reliability and maintainability across the team.
Miao Yang is currently an AI Intern at OpusClip, with an inferred three years of experience. Academically, Miao Yang pursued Artificial Intelligence Systems for a Master's degree at the National University of Singapore and holds a Bachelor's in Artificial Intelligence from Donghua University. Their tinkerer role is listed as AI Engineer (Application Development). Miao Yang is located in Shanghai, China.
none
no projects mentioned
Chen Zihao RSVP Approved
Master of Technology at National University of Singapore
designed and implemented the multi-agent architecture, including the DirectorOrchestrator, StoryLoaderAgent, StoryboardAgent, FilmCrewAgent, and VoiceActorAgent. They built the dependency-aware execution model that enables parallel generation of script, storyboard, audio, and video. They integrated Gemini Pro 3 for agent reasoning and Veo 3 for video generation, and developed the backend orchestration logic that manages routing, prompt-chaining, and inter-agent communication. They also implemented the final MP4 assembly pipeline and ensured end-to-end reliability.
Zihao Chen, whose name is also given as Chen Zihao, is currently a Master of Technology student at the National University of Singapore. They are seeking full-time work in the role of an AI Engineer, specializing in Machine Learning and Models.
LLM, Computer Vision
Current projects include DINOv2-based image retrieval (Dinov2-Retrieval, Dinov2_Matching), depth-vision robotics grasping (DummyGrasp) using YOLOv8 and Intel RealSense, and a quantitative trading prototype (Alpha-Trade) built with a FastAPI/React stack. This work involves PyTorch, Python, C++, and end-to-end ML pipelines.