ML/AI Engineer

Bridge Neurotech, Inc. is building ML capabilities to support a neural decoding pipeline that moves from functional ultrasound acquisition through reconstruction, preprocessing, and decoding. In this role, you will own and scale key parts of the modeling and infrastructure with an emphasis on real-time, low-latency inference, working closely with imaging and neuroscience teams to turn experimental findings into reliable systems.

What you will do

• Design, implement, train, and evaluate models across the pipeline, including fUSI reconstruction, signal preprocessing, and downstream decoding, and take ownership of results for the components you lead.
• Drive the experimental agenda for your area by forming and testing hypotheses about learning failures, running ablations, and tuning architectures, losses, and training schedules to produce actionable decisions for the team.
• Extend models to new data regimes, such as shifting dimensionalities (for example, 2D to 3D volumetric), new probes, and new subjects, including building preprocessing pipelines to normalize data for training.
• Own training infrastructure including experiment tracking, config management, checkpointing, mixed-precision and multi-GPU training, and profiling that others can rely on.
• Define evaluation strategy with metrics, pre-launch tests, and regression checks to confirm when models are truly improving.
• Debug numerical and performance issues such as NaNs under mixed precision, silent shape bugs, page-cache thrashing, dead gradients, and GPU memory limits.
• Build inference paths for full-volume reconstruction and low-latency real-time decoding.
• Collaborate with imaging scientists and neuroscientists to interpret data origins and artifacts, and help inform how future data is collected.
• Stay current with relevant literature in neural operators, neuroimaging, and neural decoding, reproduce key results, and recommend next steps.
• Review code and designs, and support onboarding for new engineers and interns as the team grows.

Qualifications

• PhD in machine learning, computer science, electrical engineering, computational neuroscience, physics, or a related field, or a bachelor’s or master’s degree in one of these fields plus 2 to 3 years of experience building and shipping machine learning systems in a research or product setting.
• Strong Python and hands-on PyTorch development (or equivalent frameworks such as JAX or TensorFlow).
• Demonstrated ability to take a model from a paper or prototype to something that trains reliably on real data.
• Evidence of owning an open-ended technical problem end to end, such as a thesis, publications, open-source work, or a shipped system.
• Comfort with large numerical datasets, for example hundreds of gigabytes to multiple terabytes, including memory-mapped arrays.
• Solid grounding in first principles including linear algebra and signal processing.
• GPU proficiency with mixed precision, memory profiling, gradient checkpointing, and identifying bottlenecks.
• Strong software engineering practices: version control, testing, code review, and coding conventions.
• Self-directed and motivated to execute on a goal.

Tech stack

• Python, PyTorch, JAX, TensorFlow

Benefits

• Excellent medical benefits
• 6% match 401(k)

Nice to have

• Experience with neural operators (FNO), implicit neural representations, or physics-informed machine learning architectures.
• Exposure to neural decoding, brain-computer interfaces, or neuroimaging data.
• Familiarity with complex-valued neural networks or spectral methods.
• Experience developing real-time or low-latency inference systems.
• Distributed or multi-GPU training at scale.
• Publications at machine learning or neuroscience venues, or contributions to open-source machine learning tooling.
• Prior work in a startup or small team, or evidence of thriving without much scaffolding.

Location: Redwood City, CA (onsite)