Team
Design Dash
Project Concept
No description has been added yet.
Entry
Status: Submitted
Last saved: May 09 at 5:58 PM EDT
Team Roster
Message board not available for this team yet.
Bijun Wu Team Lead RSVP Approved
Student at Northeastern University
Project architecture and MCP server design. Built the initial Gemini-based tool pipeline (generate-landing-page, update-section, regenerate-page). Designed the widget component structure (SectionRenderer, control panel layout). Coordinated the overall hackathon strategy and submission.
Bijun Wu is a Teaching Assistant at Northeastern University with 3 years of experience. They are pursuing studies in Artificial Intelligence and Computer Science at Northeastern University and Rensselaer Polytechnic Institute. Bijun is open to new opportunities and interested in technical architecture and knowledge sharing.
Harit Tarwani RSVP Approved
Teaching Assistant (DS4300) / Masters Student at Northeastern University
Implemented the agentic feedback loop — the bidirectional widget-to-server communication using mcp-use's callTool and sendFollowUpMessage APIs with cascading fallbacks. Built the refine-section and analyze-page tools on the server. Added global styles, section reordering, per-section refine bars, the prompt bar, and the agent activity log. Migrated the backend from Google Gemini to Groq API (Llama 3.3 70B) using raw fetch with JSON object mode. Used the Manufact mcp-use React widget SDK and Groq API.
CS master's student at Northeastern (Khoury). Currently TA for DS 4300 (Large-Scale Storage & Retrieval). Built a desktop calendar app in Java with MVC, Swing GUI, 8 design patterns, and 1,700+ tests in CS 5010 PDP. Building Kazam, a gamification backend in Java 21/Spring Boot with domain-driven design. Previous research internship at Academia Sinica — re-engineered a data pipeline from Python to C/C++ for a 64% throughput gain on 10M+ records. Ships code, not slides.
Generative UI — how agents render interactive components instead of returning text. Agentic tool-use loops where the UI is both the output and the steering mechanism. Interested in CopilotKit and AG-UI for connecting agents to React frontends. Also exploring how MCP apps can expose backend APIs as interactive hooks. Looking to connect with engineers building agent-driven workflows beyond chat, especially anyone working on latency-optimized rendering or real-time agentic feedback loops.
Building a generative UI layer for Kazam, a gamification platform. Instead of merchants filling static forms, an agent interprets intent ("launch a 20% discount campaign this weekend") and generates the full setup interface — dynamic fields, reward tiers, budget sliders, live scratch card preview — from its reasoning. Stack: React + CopilotKit for the copilot layer, AG-UI for transport, Java/Spring Boot backend with 17+ REST endpoints as tool-enabled hooks the agent calls directly.