# Engineering Agentic Workflows: A Full-Stack Perspective

> Three production-focused talks on engineering agentic workflows: inference loops, agent skills for full-stack dev, and real-world architecture.

This event has ended.

- When: Tue, May 12, 2026 18:00–21:00 JST
- Where: Shibuya, Japan · In person
- Region: Kanto (Tokyo)
- Organizer: Tokyo AI
- Language: EN
- Source: Luma — https://lu.ma/zn7pp9rj
- Page: https://japantech.info/events/luma-zn7pp9rj/

## Summary

An in-depth technical evening on building production-grade agentic systems, with three back-to-back talks. Shekhar Upadhaya (Skyhost) opens with the fundamentals of inference loops and patterns for long-running agent tasks, walking through a real multi-file refactor. Kaustubh Hiware (Mercari) follows with how to use structured agent skills as building blocks for end-to-end full-stack development, including designing your own skills and self-improving memory patterns. Jeff Yang (Noeon Research) closes with a strategic framework for architecting domain-specific agents in production — comparing scaffolding, workflow patterns, and flexible harnesses, and weighing RAG vs. fine-tuning. Networking follows from 20:00 to 21:00.

## About the community

A technical meetup for engineers and architects working on agentic AI, hosted by Tokyo's largest AI community (4,000+ members of engineers, researchers, investors, and PMs).

#ai #agents #llm #full-stack #architecture #production #tokyo #agentic-workflows #production-ai #rag