AI engineering · Agents · Engineering leadership

Ofer Shapira

Building AI systems. Leading the teams that use them.

I am Ofer Shapira, an AI Engineering Team Lead at Elementor, a developer and a former SaaS founder. My work sits between software, product and people: building AI systems, leading engineering teams and testing how agents change day-to-day work.

This blog is my professional writing home. I document what we built, what we tried, what failed and what we changed. It brings together my LinkedIn posts, guides and essays so the work stays useful beyond the feed.

What I focus on

• AI in software delivery: coding agents, task definition, code review and checking the result.
• Agents connected to real systems: tools, MCP, automation and context management.
• Engineering leadership: AI adoption, responsibility for decisions and the shift from writing code to judging it.
• Developer tools and open source: focused tools that remove repetitive work and make ideas testable.

Background

I have more than 15 years of software engineering and leadership experience. Before Elementor, I worked at monday.com, led engineering teams at Via and Novidea, and founded DSM Tool, a SaaS business later acquired by AutoDS. I hold a B.Sc. in Computer Science and an MBA, and served as a software developer in Unit 8200.

That background shapes the questions I ask: does the system solve a real problem, how does it fit the team's work, and how do we know it works after the demo?

Open projects

• real-browser-mcp - connect agents to a real Chrome session with existing logins.
• agents-control-tower - launch and monitor Cursor agents from the terminal.
• cursor-usage-tracker - track a team's Cursor usage and cost.

Start reading

These first-hand examples are in Hebrew; the archive also includes selected translations.

• Make code review easier with a PR description - four parts that help explain a change before reading its code.
• Keep a team's environment variables in sync - replace manual coordination with a step at application startup.
• Keep agent context files accurate - distinguish a broken reference from a problem worth fixing.

Contact and profiles

LinkedIn - writing and professional contact
GitHub - projects and code

This is my personal writing. Historical posts reflect when they were written; imported posts link to their original source.