Research Scientist, Machine Learning (PhD)

Synaptrix Labs•New York, NY

About The Position

Synaptrix is building non-invasive brain-computer interfaces by treating neural decoding as a fundamental machine learning problem. The brain produces extraordinarily high-dimensional, noisy, non-stationary signals generated by an underlying dynamical system that we can only partially observe. We are developing new models, datasets, and hardware to learn these dynamics and translate them into real-time control of computers, communication systems, mobility devices, and eventually a much broader class of machines. We are looking for exceptional researchers across machine learning, artificial intelligence, applied mathematics, physics, dynamical systems, computational neuroscience, and related fields. Prior experience in neuroscience or brain-computer interfaces is not required. We care much more about exceptional research ability, mathematical depth, and the ability to develop new approaches to difficult modeling problems.

Requirements

PhD or equivalent demonstrated research ability in machine learning, computer science, applied mathematics, physics, statistics, computational neuroscience, electrical engineering, or a related technical field.
Evidence of exceptional ability to conduct original research.
Strong mathematical foundations in areas such as linear algebra, probability, optimization, statistics, information theory, or dynamical systems.
Strong programming ability and experience implementing and evaluating machine learning models in PyTorch, JAX, or equivalent frameworks.
Experience working with high-dimensional, sequential, scientific, sensory, or otherwise complex datasets.
Ability to take an ambiguous research problem from first principles through formulation, experimentation, analysis, and implementation.
Ability to operate independently, question existing assumptions, and pursue technically ambitious ideas.

Nice To Haves

Representation learning and self-supervised learning
Foundation models
Generative modeling
Time-series or sequence modeling
Latent-variable and state-space models
Dynamical systems and system identification
Scientific machine learning
Inverse problems
Reinforcement learning and optimal control
Information theory
Statistical physics
Computational neuroscience
Neural signal processing
Multimodal learning
Large-scale distributed model training

Responsibilities

Develop new machine learning methods for modeling high-dimensional neural and behavioral data, spanning representation learning, generative modeling, sequence modeling, latent-variable models, and learned dynamical systems.
Learn latent structure and dynamics from noisy, non-stationary, partially observed time-series data.
Develop approaches to neural decoding that generalize across people, sessions, tasks, and recording conditions.
Explore problems at the intersection of deep learning, dynamical systems, system identification, control, information theory, optimization, and statistical learning.
Investigate self-supervised and unsupervised learning methods that can take advantage of large quantities of neural data without requiring dense behavioral labels.
Design rigorous experiments to understand model scaling, generalization, representation quality, and the limits of non-invasive neural decoding.
Build simulations and generative models for studying neural signals and testing hypotheses about learned representations and decoding algorithms.
Work closely with researchers collecting large-scale neural datasets and engineers building the sensing hardware that generates them.
Translate promising research into real-time systems controlling computers, communication interfaces, wheelchairs, prosthetics, and other machines.
Build rigorous, reproducible implementations of research ideas and scale successful approaches to large datasets and compute.
Contribute original research that advances both Synaptrix's systems and the broader scientific understanding of neural decoding.