turboseti served as the workhorse software and algorithm used to search for extraterrestrial intelligence over the the past decade. The Breakthrough Listen project has an updated pipeline for narrowband doppler drifting signals that learns from the last decade of data collection from the Green Bank Telescope, The Allen Telescope Array, Very Large Array, and Meerkat.

This talk will give an overview of the new and improved narrowband doppler drifting search pipeline created for the Breakthrough Listen project which searches for signs of extraterrestrial intelligence using radio telescopes. The new pipeline called BLISS (Breakthrough Listen Interesting Signal Search) includes a new C++-based device-aware ndarray library called BLAND (Breakthrough Listen Arrays with N-Dimensions) heavily utilizing dlpack for easy interaction with a large ecosystem of neighboring scientific computing tools used in software radio such as numpy, cupy, matplotlib, pytorch, tensorflow. We will show how this library works with a quick overview of "why ndarrays" and build on top of that to work through the whole pipeline from 100m dish to detected signals.

The current pipelines process hundreds of MHz of instantaneous bandwidth which is channelized in to various data products including down to channels of single-digit Hz. These fine channels are then used to search for narrowband doppler drifting signals. This search is sorting through the galactic haystack for needles which don't originate from earth. We'll highlight how we mitigate radio frequency interference, algorithmically search for signals with both single dish & arrays, what's new & corrected in the updated pipeline from historic SETI pipelines, and ways to join the SETI open source community.

In addition to the focus of narrowband doppler drifting signal search with device-aware ndarrays that provide absurd ecosystem compatibility, we'll show off polyphase channelization and signal processing with ndarrays that demonstrate how BLAND can be used generally for software radio.