Data Scientist, Biosignal Deep Learning

Naqi Logix builds neural earbud and micro-gesture interface technology that interprets physiological signals as digital intent. The Data Scientist, Biosignal Deep Learning will design, train, evaluate, compress, and deploy deep learning models for biosignal time series, collaborating with firmware, hardware, and research teams while mentoring junior colleagues.

Responsibilities:

• Design and iterate on architectures for detecting gestures, intent, and states from multichannel biosignal streams. These include 1D CNNs, temporal convolutional networks, and lightweight hybrid models
• Own experiments from start to finish, from the first hypothesis through data preparation, training, evaluation, and documentation
• Review existing model and signal-processing code and lead improvements to it, including the detection pipelines for jaw-clench and head-gesture input
• Design preprocessing pipelines: filtering, artifact handling, windowing, normalization, and feature extraction suited to each signal type
• Build evaluation methods that fit the realities of biosignal data. These include differences between subjects, drift across sessions, differences in electrode contact and fit, and leakage between training and test sets
• Compress and deploy models for edge inference within real limits on latency, memory, and power. You will use quantization, pruning, and efficient architecture choices, and you will work with firmware engineers on how models are integrated
• Turn synchronized multi-device validation study data into training and benchmark datasets
• Use Claude Code and other frontier coding tools to speed up exploration, testing, and refactoring. Review AI-generated code as carefully as you would review a colleague's pull request, and help define how the team does this
• Document experiments so they can be reproduced and compared, and mentor junior team members as the team grows