Research Scientist: Hierarchical Sensorimotor Perception

Astera Institute; Emeryville (CA); On-site; USD 90,000 - 130,000; Full time

Job summary

An innovative research institute in Emeryville, California seeks a Research Scientist to develop foundational principles of Artificial General Intelligence (AGI). Responsibilities include creating new bi-directional architectures for perception, conducting key experiments, and collaborating with interdisciplinary teams. Ideal candidates have a PhD in a quantitative field and robust skills in deep learning frameworks like PyTorch. Join us to tackle the big questions of AI in a mission-driven, collaborative environment.

Qualifications

• PhD or exceptional candidates in related quantitative fields.
• Strong foundations in deep learning, graphical models, and information theory.
• Expert-level coding skills in deep learning frameworks.

Responsibilities

• Develop new bi-directional architectures for hierarchical perception.
• Design experiments for testing generalization against existing architectures.
• Collaborate with scientists and engineers on scalable systems.
• Innovate new publication models for open science.

Skills

Deep learning

Graphical models

Information theory

Python

PyTorch

JAX

Education

PhD in Computer Science, Electrical Engineering, Neuroscience, Physics

Job description

Astera Institute

Research Scientist: Hierarchical Sensorimotor Perception

About the Role

We are seeking brilliant, unconventional thinkers to join us. In this role, you will not be chasing incremental gains on standard benchmarks. Instead, you will be tasked with developing the foundational principles of Artificial General Intelligence. We believe that understanding the "code of the brain" is the most viable path to building truly intelligent machines, and we are looking for scientists who can bridge the gap between biological intelligence, computational theory, and building at scale.

Core Responsibilities

• Fundamental Research on Hierarchical Perception: Develop and implement new bi-directional/recurrent architectures for hierarchical perception. These models are expected to provide flexible querying, dynamically variable levels of abstraction, and representations that support planning/reasoning.

• Evaluation and Benchmarks: Design experiments to test and benchmark against existing architectures for out-of-distribution generalization.

• Cross-Disciplinary Collaboration: Work alongside neuroscientists and software engineers to translate abstract mathematical frameworks into scalable systems.

• Contribute to Open Science: Help innovate new publication models that incentivize speedy dissemination, open source code releases, free open access, and impact measurements based on uptake.

Desired Qualifications

• Technical Depth: Preferred: PhD in Computer Science, Electrical Engineering, Neuroscience, Physics, or a related quantitative field. We are always open to considering exceptional candidates, even those with non-traditional educational histories.

• Theoretical Rigor: Strong foundations in deep learning, graphical models, and information theory.

• Coding Proficiency: Expert-level skills in deep learning frameworks (PyTorch/JAX), with a focus on clean, reproducible research code.

• AGI Mindset: A demonstrated interest in the "big questions" of AI—robustness, generalization, and common-sense reasoning.

• Curious About the Brain: A genuine passion for exploring the computational and architectural principles of the mammalian brain.

• Startup DNA: Ability to thrive in a lean, fast-paced environment where you have high autonomy and a direct influence on research direction.

Why Join Us?

• We recognize the need for fundamental research: No pressure to align research with quarterly revenue goals, market competition, or product roadmaps. We can focus on solving the core problems and achieving breakthroughs that enable the next generation of AGI technology.

• Mission-First: Your work contributes directly to the global pool of knowledge and the safe development of AGI.

• Collaborative culture: We work in teams, organized around challenge problems and attack paths, without big-company inertia or politics.