​Date: Wednesday, July 8th, 2026
Time: 18:00 to 20:00
Location: Tikal Offices

AI agents are moving from demos into real engineering workflows. Once they need to plan, call tools, coordinate steps, and keep context across longer tasks, the challenge becomes less about prompting and more about system architecture.

​This meetup looks at two connected layers of that challenge.

​Rivka Spiegel will start with the broader architecture view: how agentic systems are structured, which patterns and frameworks are emerging, and how teams can design systems that are easier to coordinate, evaluate, and control.

​Amit Goldstein will then dive into one of the most critical layers inside these systems: memory. He will show how agents can keep context over time instead of starting from zero every time, and what it takes to make that memory useful, reliable, and practical.

​On the Agenda

​18:00 to 18:30 → Welcome Drinks & Networking

​18:30 to 19:15 → Agent Architectures and Frameworks// Rivka Spiegel, Senior Backend and Data Engineer, Tikal
*The talk will be held in English

​As agentic systems grow, architecture becomes the main challenge. A single prompt or agent loop may work for a limited task, but production grade workflows require clearer patterns for planning, coordination, correctness, and human oversight.

​In this session, Rivka will examine the shift from simple LLM prompting to robust agentic system design. She will walk through key design patterns for 2026, including Single Agent ReAct loops, sequential workflows, and hierarchical agent teams using shared RAG and Human in the Loop checkpoints.

​The session will also cover the current framework landscape, including tools like LangGraph for fine grained workflow control and AG2 for multi agent conversations. The goal is to give developers a practical view of how to decompose tasks, coordinate agents, and design systems that can operate with more structure in real environments.

​19:15 to 20:00 → Agentic Memory In Practice// Amit Goldstein, Senior Backend Engineer, Tikal
*The talk will be held in Hebrew

​LLMs are powerful, but by default they suffer from digital amnesia. Every interaction starts fresh, without the deep and personalized context needed for autonomous behavior over time.

​In this session, Amit will explore the shift from reactive prompting to persistent agentic memory. He will compare current architectural approaches for memory and context management, and then dive into a live coding demo using Mem0.

​The talk will focus on how memory changes the way we design agentic systems, what it enables, and where teams need to be careful when adding persistence, personalization, and context into AI workflows.

​Please Note

• ​Registering through this page is an application for approval and will be reviewed by the team before tickets are confirmed.

• ​By registering for this event, you agree to receive emails from Tikal.

​About Tikal

​Tikal is a hands-on tech consultancy partner for scaling engineering organizations and creating meaningful tech impact. With over 25 years of experience, we work side by side with engineering teams across AI and ML, Backend, Data, DevOps, Fullstack, and Web.

​As the creators of the Israeli Tech Radar, we actively share knowledge and insights from the field to help engineering leaders and teams make better technology decisions.