Yair Alon, who developed key deep learning models like Portrait Mode at Google, has transitioned to the startup world as a founding research scientist at Poetiq, a company developing AI that can optimize itself and other AI systems.
- Poetiq focuses on AI that improves itself and other AI systems.
- Yair Alon builds experimentation infrastructure from scratch at Poetiq.
- Startup culture offers challenge, autonomy, and broad impact opportunities.
What happened
Yair Alon, a veteran researcher who contributed to Google’s Portrait Mode and other AI projects at DeepMind, left Google after 10 years to join Poetiq, a new startup focused on developing an AI system capable of optimizing other AI systems as well as itself. Alon is a founding research scientist at Poetiq, where his early work involved building novel experimentation infrastructure that supports rapid validation of their recursive optimization technology.
At Poetiq, Alon is deeply involved in all aspects of the startup’s growth, from developing machine learning techniques to handling operational tasks, embodying the versatile role early employees often take on. The company’s core innovation lies in recursive self-improvement, an ambitious approach that aims to accelerate AI progress by enabling systems to enhance their own capabilities autonomously.
Why it matters
Alon’s transition from a stable corporate research setting to a startup reflects a broader trend where tech innovators seek environments that facilitate faster, more flexible advancements. Poetiq’s pursuit of AI that can self-optimize could significantly impact how AI systems evolve, potentially increasing efficiency and reducing the necessity for constant human oversight in AI development.
The project’s dual focus on foundational AI algorithms and bespoke experimentation infrastructure highlights the challenges of creating truly novel AI products outside large organizations. Poetiq demonstrates how startups with deep technical expertise can push boundaries by combining academic rigor with rapid product iteration and cross-functional collaboration.
What to watch next
Poetiq’s progress in scaling their recursive self-optimization technology will be a key indicator of its potential impact in AI research and industry applications. Observers should watch how effectively the startup manages to run increasingly complex experiments with minimal human intervention and whether it can accelerate AI capability improvement at scale.
Additionally, Poetiq’s hiring and team-building efforts will shape its future innovation velocity. As the company expands, maintaining the autonomy and collaborative culture praised by Alon will be important to sustaining the creativity needed to succeed in this new AI frontier.