ZyG's $60M Plan: Hire 50 Engineers to Build Multi-Agent Infrastructure

ZyG raised $60 million from Accel at a $500 million valuation this week, just two months after emerging from stealth. The twist: the five IronSource co-founders aren't building another e-commerce tool. They're creating what they call an "Agentic Operating System."

The company plans to hire 50 engineers in the next six months, with a structure unlike typical AI startups. Instead of product teams, ZyG is organizing around agent capabilities: perception, planning, execution, and memory.

The Agent-First Org Chart

Each capability gets its own engineering lead and dedicated team. The perception team builds agents that understand business context from Shopify data, ad performance, and customer behavior. Planning engineers create agents that set strategy across channels. Execution teams handle the actual campaign management and inventory decisions. Memory engineers ensure agents learn from outcomes.

"We're not hiring ML engineers to tune models," a source familiar with the hiring plans told me. "We're hiring systems engineers who can make agents work together."

The roles reflect this focus. ZyG posted openings for "Multi-Agent Orchestration Engineers" at $285,000 base salary. "Agent Reliability Engineers" command $265,000. Traditional titles like "Senior Backend Engineer" don't appear in their job listings.

Why Multi-Agent Architecture Matters Now

ZyG's platform already manages advertising, customer support, and inventory for direct-to-consumer brands. But unlike single-purpose AI tools, their agents operate autonomously across functions. An inventory agent might notice a stockout risk and tell the advertising agent to slow campaign spend. A support agent spots a product complaint trend and alerts the product team.

This requires infrastructure that doesn't exist yet. Current frameworks assume one agent per task. ZyG needs agents that negotiate with each other, share context, and make joint decisions.

Accel partner Rich Wong noted the technical challenge: "Even for those brands that successfully achieve early traction, the path to scaling remains challenging." The bet is that multi-agent systems can handle that complexity better than humans managing separate tools.

The Hiring Challenge

Finding engineers who understand distributed systems and agent behavior is tough. ZyG isn't competing with OpenAI for researchers. They're competing with Uber and Airbnb for engineers who've built marketplace orchestration.

The company is targeting engineers from three pools:

1. Robotics veterans who've worked on multi-robot coordination 2. Gaming engineers who've built NPC behavior systems 3. Trading systems engineers who understand autonomous decision-making

One recruiter working with ZyG described the search: "We're looking for people who've made independent systems cooperate. That could be robots in a warehouse or bots in a game."

Building the Bench

ZyG's technical leadership reflects this interdisciplinary approach. The five IronSource founders bring mobile ad-tech expertise. They're now hiring directors from outside e-commerce: a principal engineer from Boston Dynamics, a tech lead from Riot Games, and a systems architect from Jump Trading.

The Series A gives them runway to build this team before revenue pressure hits. At current burn rates, they have 18 months to prove multi-agent architecture works at scale.

The first test comes in Q3 when ZyG launches its multi-agent SDK. If they can show agents making complex decisions across business functions, it validates the hiring strategy. If not, they've built an expensive engineering team for a problem that might not exist.

For builders considering similar architectures, ZyG's org structure offers a blueprint: organize around agent capabilities, not product features. Hire for systems thinking, not ML expertise. And give agents room to surprise you.