Research Engineer - ML Systems & Data Pipelines

constellation; San Francisco (CA); On-site; USD 150,000 - 190,000; Full time

Benefits offered by this job

Health, dental, vision insurance

Relocation assistance

Office in SF Mission District

Team dinners & off-sites

Parental leave

Job summary

Constellation in San Francisco is seeking a Research Engineer to sit between data and models, orchestrating and optimizing training on long-horizon multimodal sequences while building pipelines to turn a growing corpus into learnable material for models.

You will write research code, align data across streams, and ensure reproducibility as the team experiments and scales the workflow with engineering and research colleagues.

Qualifications

• 3+ years building ML systems or research infrastructure, including distributed training runs on 100+ GPUs.
• Expert-level Python and deep knowledge of PyTorch internals: DDP/FSDP, mixed precision, gradient accumulation, profiling tools.
• Track record of building or maintaining research libraries others depend on.
• Experience with data pipelines over large unstructured and multimodal datasets and familiarity with columnar/streaming formats.
• Hands-on experience with experiment tracking and dataset versioning tooling.
• Ability to read papers, reimplement components, and evaluate loss curves critically.
• Strong debugging skills across shards, collate functions, and training runs.
• You thrive in a high-bandwidth, collaborative environment and care about the impact on people.

Responsibilities

• Orchestrate and optimize training: distributed config (DDP/FSDP), mixed precision, checkpoints and profiling.
• Build high-throughput data loading over TB-scale multimodal data: sharding, prefetching, caching and format choices.
• Wrangle complex data: align/synchronize multi-stream time series; manage dataset versioning.
• Derive rich features from raw recordings and version datasets for reproducibility.
• Write clean, tested research libraries for datasets, models, transforms and metrics.
• Make runs traceable: code version, config, dataset version and environment.
• Build evaluation harnesses that surface regressions on new checkpoints.
• Turn research prototypes into repeatable pipelines while preserving exploratory flexibility.

Skills

ML systems

Python

PyTorch internals

Distributed training

Experiment tracking

Data pipelines

Research libraries

Tools

torch_geometric

torchaudio

torcheeg

torch_brain

neuralsets

Job description

Constellation in San Francisco is seeking a Research Engineer to sit between data and models, orchestrating and optimizing training on long-horizon multimodal sequences while building pipelines to turn a growing corpus into learnable material for models.

You will write research code, align data across streams, and ensure reproducibility as the team experiments and scales the workflow with engineering and research colleagues.