Neural Data Infrastructure Engineer
Owns the end-to-end data infrastructure that turns intracortical neural recordings into reliable, training-ready datasets for Blackrock Neurotech's neural foundation models — building ingestion, preprocessing/validation, storage, and GPU-feeding data loaders. Core stack: Python, neural signal processing (spike sorting), scientific formats like NEV/NSx, NWB, HDF5 and Zarr, plus PyTorch data interfa
- Compensation
- Not disclosed
- Location
- Salt Lake City, UT
- Work setup
- Onsite
- Seniority
- Not specified
- Employment
- Not specified
- Experience
- 1+ years
- Category
- Devops
- Posted
Skills
Must have
- Demonstrated experience independently designing and operating large-scale scientific data pipelines end to end, with ownership of software quality, reliability, and delivery
- Bachelor's degree in Computer Science, Electrical Engineering, Biomedical Engineering, Neuroscience, or related technical field, or equivalent practical experience
- Exceptional programming ability in Python, with strong software design, debugging, profiling, automated testing, and version control practices
- Advanced understanding of extracellular neural recordings, spike detection and sorting, sampling theory, filtering, aliasing, referencing, and artifact handling
- Hands-on experience with intracortical electrophysiology data and its on-disk representation (binary layouts, event timestamps, channel metadata, continuous signals)
- Ability to reason about how preprocessing choices alter neural information and affect model training, evaluation, and reproducibility
- Experience with scientific formats and storage systems such as NEV/NSx, NWB, HDF5, and Zarr, including efficient partial reads and metadata management
- Strong judgment about dataset lineage, validation, access permissions, and handling of sensitive research data
- Ability to work across neuroscience, machine learning, and infrastructure teams and bring clarity to ambiguous requirements
- On-site at Blackrock Neurotech headquarters in Salt Lake City, Utah; occasional travel may be required
Nice to have
- Experience with distributed or parallel processing, object storage, workflow orchestration, and restartable jobs at scale
- Familiarity with deep learning models and architectures, including how data processing decisions affect model training, inference, and evaluation
- Experience processing local field potentials, synchronizing neural and behavioral streams, or supporting longitudinal neural datasets
- Experience with compiled languages, high-throughput I/O, cloud cost management, or performance-sensitive scientific computing
- Fluency with deep learning data interfaces, including tensor shapes, batching, masking, sampling, and PyTorch
Full description
The Role
The Neural Data Infrastructure Engineer will own the systems that turn intracortical recordings into reliable, training-ready data for AI/ML models. You will build the data infrastructure needed to meaningfully scale model capability while preserving the scientific meaning of every recording.
As a hands-on individual contributor on a small research team, you will be the technical authority on the data lifecycle from upload through training consumption. You will work closely with model researchers to understand their data requirements and translate them into reliable, scalable infrastructure. You will also partner with infrastructure and IT teams to ensure the right storage, processing, and delivery resources are in place. You will take end-to-end technical ownership of a growing intracortical BCI data corpus that sits at the foundation of Blackrock's neural foundation model work. Your work will directly shape the quality, scale, and reliability of the data these models learn from, while giving researchers the confidence and freedom to focus on model development
What You'll Do
- Own the design, implementation, and operation of the neural data pipeline from upload and validation through preprocessing, dataset releases, and training data loaders
- Build reliable ingestion for heterogeneous recording formats, preserving raw data, acquisition metadata, channel and electrode mappings, timestamps, units, and behavioral alignment
- Implement and validate neural signal processing for spike events and waveforms, including detection or sorting where needed, quality assessment, binning, and normalization; extend the pipeline to field potentials as the program evolves
- Establish automated checks for corrupt files, missing channels, clock drift, artifacts, recording discontinuities, and inconsistent metadata, with clear criteria for quarantine and recovery
- Design storage layouts, indexing, chunking, compression, and caching that support efficient access to large neural datasets across local and cloud resources
- Create reproducible, versioned datasets with traceable transformations, provenance, and subject and session partitions that prevent leakage between training and evaluation
- Partner with model researchers to define input representations, sequence construction, masks, sampling policies, and interfaces that preserve neural meaning and meet architecture requirements
- Build and profile parallel preprocessing and streaming data loaders, partnering with the training performance engineer to keep GPUs supplied as training scales
- Work with infrastructure, IT, and data owners to plan capacity, access controls, retention, backup, and recovery appropriate to the recordings and their approved uses
- Establish testing, monitoring, documentation, and operational practices that make pipeline failures diagnosable and dataset releases dependable