- Hemlock
- San Francisco, CA
- Full-Time
- 51 days ago
- $50,000 – $100,000
Software/ML/Signal Processing Engineer.
Before you go
Before you leave us, sign up for our email alerts
We don't do job spam, just the best digital jobs delivered straight to your inbox.
Software/ML/Signal Processing Engineer: our view in 3 lines...
- The Role:This role is for a software, machine learning, and signal processing engineer building the production software stack for a wrist-worn surface EMG device.
- The Person:The person will build the production software stack from raw analog signal through deployed inference, including the signal processing stack, data transfer protocols, API, and client dashboard.
- Requirements:The advert asks for software, ML, and signal processing work, plus a production system, data transfer protocols, API, client dashboard, and full stack delivery.
About the role
You will own everything between the electrode and the API. Hemlock has working hardware, a synchronized multimodal data collection system, and a locked model architecture. What it does not have is a production system. That is the job.
This is a founding technical role reporting directly to the CEO. You will be the second technical decision-maker in the company and the first person whose primary output is software that ships to customers. Signal processing stack, data transfer protocols, API, client dashboard, the works. If you want a narrow research problem handed to you with the infrastructure already built, this is the wrong role. If you want to own a full stack from raw analog signal to deployed inference and have your architectural calls actually stand, it is the right one.
Every serious robotics lab is bottlenecked on the same thing: demonstration data. The field has converged on egocentric video and hand pose as the substrate, and both are now cheap to collect at scale. What neither of them contains is force. A video of someone picking up an egg and a video of someone picking up a rock are nearly identical in pixel space and nearly identical in joint angles. They differ almost entirely in a quantity nobody is recording. Policies trained without it learn geometry and inherit no sense of touch.
Hemlock closes that gap at the source. Our device, the Bracelet, is a wrist-worn surface EMG band that reads the electrical activity of the forearm muscles that actually drive the hand, which means it recovers contraction force directly rather than inferring it from motion. This distinction is the whole thesis: force cannot be reconstructed downstream of kinematics, because a hand squeezing hard and a hand resting can occupy the identical pose. Muscle signal is the only non-invasive channel that separates them.
