We are looking for someone to own the question at the center of Subvocal: what can we physically measure from the human body that contains enough information to recover internal articulation?
The broader technical design space includes RF sensing, EMG, EEG, mmWave, and other non-invasive physiological sensing methods. For competitive and IP reasons, we are not publicly disclosing the exact architecture of our current system yet, although we share much more during the interview process.
The signals we care about are extremely subtle. They are affected by anatomy, sensor geometry, device placement, motion, interference, and changes of only a few millimeters. A sensing configuration that looks excellent on one person can fail on another, and a configuration with the highest apparent signal strength is not necessarily the one that contains the most useful information for decoding language.
You will own the sensing system from first principles through a wearable implementation. That includes deciding what measurements matter, designing the experiments that answer those questions, and working closely with ML to evaluate configurations based on actual cross-user decoding performance rather than isolated signal metrics.
Some of the problems you will work on include:
•
Designing and evaluating new sensing geometries, modalities, channels, and frequency configurations.
•
Understanding how anatomy and articulator movement affect the measured signal.
•
Improving signal-to-noise ratio while preserving the information needed to distinguish similar phonemes.
•
Building multi-channel systems and determining what genuinely independent information each channel adds.
•
Modeling and measuring the interaction between sensors, electronics, mechanical design, and the body.
•
Developing calibration and normalization methods that make measurements comparable across people, sessions, and devices.
•
Translating a laboratory sensing setup into a compact, low-power wearable.
•
Working with external RF, fabrication, simulation, and regulatory partners where useful.
You might be a great fit if you have deep expertise in RF engineering, antennas, radar, electromagnetics, biomedical sensing, applied physics, signal processing, or a related area. We are especially interested in people with a PhD or equivalent research depth who are also extremely hands-on. You should be comfortable moving between simulation, mathematical reasoning, benchtop experiments, custom hardware, and data analysis.
This is not a pure simulation or advisory role. You will spend a lot of time building things, testing them on real people, discovering that reality does not match the model, and deciding what experiment to run next. You will also help build the sensing team around you and shape the fundamental architecture of the product.
This is a full-time, in-person role in San Francisco.
At a high level, we are trying to detect the incredibly subtle physiological changes that happen when someone forms words internally, and turn those signals into continuous language. The broader sensing design space includes RF sensing, EMG, EEG, mmWave, and other non-invasive methods. For competitive and IP reasons, we are not publicly disclosing the exact architecture of our current system yet, although we can share much more during the interview process.
The hard part is not just getting a model to work on one person in one recording session. These signals can change when the device moves slightly, when the same person comes back the next day, or when you move to someone with completely different anatomy. We need the system to work across people, devices, and environments, then adapt to a new user from only a few minutes of calibration. Solving that involves a mix of signal processing, self-supervised learning, large temporal models, personalization, and language decoding. We currently experiment with architectures including Conformers, Mamba-style sequence models, and pretrained speech and language models. Most of our ML stack is built in Python and PyTorch, with custom infrastructure for data collection, signal processing, distributed training, and evaluation.
At the same time, we are taking a research system built from laboratory equipment and turning it into a wearable. That means building custom sensing and mixed-signal electronics, embedded and FPGA systems, custom ASICs, high-speed data acquisition, firmware, power systems, and eventually fitting everything into a small wearable. Sensor placement, electrical design, mechanical design, and the model are all tightly connected. Moving something by a few millimeters or changing part of the electronics can alter the data distribution enough to affect the model.
A large part of our next phase is also a data problem. We are building the infrastructure to collect thousands of hours of high-quality subvocal speech data across thousands of people, train models continuously as that dataset grows, and understand exactly how performance scales with more users and more variation.
This is what makes the work unusually interesting. There is no established playbook, and the important breakthroughs can come from a better model, a better sensing configuration, a clever piece of hardware, or simply discovering that we have been framing the problem incorrectly. The people joining now will have room to work across those boundaries and make decisions that directly determine whether the system works.