Machine Vision Scientist in Stanford

n \n

• Develop and maintain machine vision pipelines for use in experimental settings. \n

• Write and test video and data analysis software and assist with linking these systems to the lab's broader experimental infrastructure. \n

• Help set up, calibrate, and troubleshoot multi-camera recording setups; support synchronization of video and other experimental data streams. \n

• Assist with analysis of experimental datasets, and help build and document clean, reproducible data pipelines. \n

• Learn and apply new computational tools as the lab's methods evolve and contribute ideas for improving existing workflows. \n

The Haroush Lab at Stanford is seeking a Postbaccalaureate Machine Vision Scientist to help build and refine tools for tracking behavior in systems neuroscience experiments. This is a full-time research position for a recent college graduate with a strong quantitative or computational background who wants to take on real technical ownership at the intersection of computer vision and neuroscience, developing skills and a publication record that carry directly into graduate school or advanced research roles., * Bachelor's degree in neuroscience, computer science, engineering, physics, or a related quantitative field, completed within the last 1-3 years. \n

• Strong Python programming skills, with experience writing new pipelines that can be integrated into experimental workflows. \n

• Hands-on experience with computer vision and deep learning - through advanced coursework, independent research, or personal projects - ideally including pose estimation, object detection, segmentation, multi-object tracking, or video analysis. \n

• Ability to configure models and environments to run on GPUs, using frameworks such as PyTorch or TensorFlow. \n

• Strong experience with data pre-processing, cleaning, and management. \n

• Ability to adapt or move beyond existing templates and frameworks when a problem calls for it. \n

• Strong written and oral communication skills, and the ability to work collaboratively in a small research team. \n, Experience working with large, multi-modal datasets (e.g., video paired with behavioral or other experimental measurements).

Prior exposure to behavioral tracking tools such as DeepLabCut, SLEAP, or Bonsai (through coursework, a lab rotation, or personal projects).

\n

Strong quantitative and debugging skills.

\n

Clear documentation and validation practices, since the resulting measurements need to be scientifically interpretable and reproducible.

\n

Undergraduate research experience in behavioral neuroscience or related experimental fields.

\n

Basic familiarity with camera hardware, video data formats, or real-time data acquisition.

\n

Coursework or project experience touching on game theory or decision-making research.

\n

\n

About the Haroush Lab

\n

The Haroush Lab studies the biological basis of complex social interactions, from dyadic interactions to group dynamics and collective decision-making. The lab seeks a mechanistic understanding of the fundamental building blocks of societies, such as cooperation, empathy, fairness, and reciprocity. Because the computations underlying social interactions are highly distributed, the lab's research asks which specific systems are involved in particular functions, why such organization arises, and how activity across multiple systems is coordinated. The long-term goal is to develop a roadmap of the social brain that can guide restorative approaches for conditions in which social behavior is impaired, such as autism spectrum disorders and schizophrenia. The lab studies social tasks based on game theory, combined with advanced analytical approaches, machine vision, and deep learning, to understand the underpinnings of complex social computations.

\n

The lab is small and highly collaborative, and postbacs are treated as active contributors to ongoing studies rather than peripheral support staff. Compensation is competitive, with full benefits provided through Stanford University.

\n

\n