Final Winner, JunctionX Korea 2026

Our team won the Grand Prize at JunctionX Korea 2026, taking both Track Winner and Final Winner awards among 250 participants after competing in the Lablup x FuriosaAI track. The award included tickets to Junction 2026 in Finland and KRW 1,500,000 in airfare support per person.

We built ARGUS, a platform for turning benchmark results into clearer agent configuration decisions. We evaluated Qwen3, GPT-OSS, and EXAONE across multiple tasks and found that keeping each task within a single reasoning context with lightweight models was simpler and more effective than unnecessarily fragmenting it across subagents.

My main focus was UI/UX, frontend development, and benchmark analysis, with additional work on agent orchestration. I built a 3D performance graph with ECharts GL, 2D benchmark visualizations with Unovis, and interactive components using React, TypeScript, and Base UI.