LayerX
Forward Deployed Engineer, LayerX
Ai Workforce事業部にて、課題設定・評価設計から本番導入まで、顧客と伴走しながらAIエージェントを構築しています。
Resume
Ai Workforce事業部にて、課題設定・評価設計から本番導入まで、顧客と伴走しながらAIエージェントを構築しています。
Activity
Why self-improving agents plateau around 90% accuracy, and a harness design that pushes past it: training data mined from execution logs, an optimizer that bounds edits, a search strategy that keeps multiple candidates, and evaluation you can trust.
Project release I contributed to — building a large-scale LLM-as-a-judge generation and evaluation pipeline.
Reached the global top 100 three times and played as an Asia representative.