Developer Relations Engineer

About Spiral

Spiral builds a fast analytical database for multimodal, multi-rate data streams, on top of the open source Vortex file format. Our users are AI/ML researchers and AI infra engineers developing models in complex domains, such as weather & climate, financial, time-series, genomics, point-clouds, videos, and images. They spend their days waiting on data loaders, writing video or sensor data pipelines, and watching expensive GPUs sit idle on I/O. We make that pain go away.

We're small, technical, and early. The way we win researchers & developers is by being useful to serious practitioners.

The Role

This is our first DevRel hire, and it is a content and developer-experience role. You'll own the material that teaches researchers and training engineers how to get real work done with Spiral: cookbooks, worked examples, benchmarks, and reference pipelines grounded in datasets people actually train on. You might also conduct research on new applications of the core Spiral product.

You'll collaborate with the rest of the team to create the tightest possible feedback loop between developers/researchers and our product and docs. The job is to earn credibility with a technical audience by shipping things that are genuinely good.

What you'll do

• Write technical cookbooks and end-to-end examples — video ingestion, GPU data loading for training, multimodal feature engineering — that a researcher can run and immediately understand.

• Build and maintain reference pipelines against real datasets (such as those on Hugging Face), and keep them working as the product moves.

• Produce credible benchmarks and the honest writeups that go with them.

• Be the developer's advocate internally: turn friction you and users hit into concrete product, client SDKs (e.g. pyspiral), and docs improvements.

• Answer real questions in the places our users already are (GitHub, community channels, conferences), and turn recurring ones into permanent docs.

• Represent Spiral at conferences.

• Give the occasional talk or workshop.

Who you are

• You've done ML or training-infrastructure work yourself, and you've personally felt the data-loading / video-decode / GPU-utilization pain we remove.

• You write well about technical things. You can point us to things you've written.

• You're fluent in Python and comfortable in the PyTorch / Hugging Face / data-pipeline ecosystem.

• You have a working mental model of GPUs, training loops, and where the bottlenecks actually are.

• You are well connected in developer and/or AI communities.