Maziar "Maz" Gomrokchi, PhD

Seeing the Age of AI Agents
Before It Arrived

Long before AI became a business buzzword, Maziar was fascinated by a simple question: Could machines act intelligently on their own?
‍
As an undergraduate computer science student, his fascination with intelligent machines led him to humanoid robotics and the RoboCup Rescue Simulation. In 2003, during the second year of his undergraduate studies, he began working in the field. By 2005, he had assembled a team of five developers to compete internationally in the RoboCup Simulation Rescue competition. They finished sixth in the world.

But robotics was only the beginning. Maziar became increasingly interested in a deeper question: how could autonomous agents make decisions in complex environments where they had to compete, cooperate, and respond strategically to one another? That led him to trading agents and the Trading Agent Competition Market Design, where autonomous software agents competed in simulated markets.

In 2007, his team finished third internationally, followed by a first-place world ranking at the competition held at the University of Michigan, Ann Arbor, under the direction of Michael Wellman and with sponsorship from Google and Yahoo.

His first major award was presented at IJCAI 2009 at Caltech in Pasadena. He also published his first academic paper. From there, Maziar’s journey expanded into the international AI research community, particularly around multi-agent systems, algorithmic game theory, and intelligent decision-making. In 2011, he started his PhD at McGill University, where he worked on quantum information theory under Patrick Hayden.

Then AI began accelerating. While attending ICML in New York in 2016, at a time when DeepMind had recently been acquired by Google and OpenAI was beginning its journey, Maziar received an opportunity to join an emerging team of AI scientists.

He chose a different path. Rather than joining the rapidly growing wave immediately, he continued working on something he believed would become increasingly important: privacy-preserving deep reinforcement learning.
His research brought him into close collaboration with the emerging deep learning community at Mila, one of the world’s leading AI research institutes, where he worked alongside top-tier researchers including Yoshua Bengio.
He continued collaborating with Google DeepMind researchers through 2020 and remained deeply involved in AI research while increasingly working with industry. Beginning in 2018, he became an AI consultant to companies and research organizations, translating advanced AI research into practical applications.
‍
Then came the moment that changed the direction of his entrepreneurial thinking. It was December 2022, at NeurIPS, when ChatGPT was publicly introduced. For most people, generative AI seemed to arrive almost overnight.

For Maziar, it was different. He had spent years watching the underlying technologies develop. He understood what was coming next. If AI systems could understand language, reason about information, use tools, and increasingly act autonomously, then eventually AI agents would become an active layer between people and the internet.

The question was no longer simply How do people find brands? It was becoming: How will AI agents find, understand, evaluate, and recommend brands?

Maziar saw a missing piece. Businesses would need a way to monitor how AI agents perceived them, understand what those systems knew about them, and communicate with those systems effectively.

That insight would eventually become one of the foundations of Elementera.

Research and publications

Writing on AI search