
Learn to reanalyze open neurophysiology data, from the people who built the tools.
A one-week residential school at HHMI Janelia on working with NWB data on the DANDI Archive: finding it, streaming it, analyzing it with community tools, and applying AI methods to it. Free to attend, including lodging and meals. Formerly NeuroDataReHack.
Applications for the 2027 school open in early 2027.

What you learn
Two strands, taught side by side, and a week of project work to put them to use.
Open data and the tools around it
How NWB files are organized and validated; how to search the DANDI Archive and stream from it instead of downloading; the flagship datasets from OpenScope, the International Brain Laboratory, and MICrONS; and the analysis ecosystem built on the standard: SpikeInterface and Kilosort, Pynapple and NeMoS, Neurosift, suite2p, Facemap, and Spyglass.
Modern methods for neural data
Foundation models for neural activity and how to apply them across datasets; coding with large language models and AI agents; dimensionality reduction, neural decoding, and state space models; and how to judge whether a method developed on one lab's recordings transfers to another's.
From the 2026 lectures
All 32 recordingsProjects from the last installment
All 84 projectsReplicating minute-scale oscillatory sequences in entorhinal cortex
Ben Dichter, Ryan Ly
Gonzalo Cogno et al. (2024) reported that mouse medial entorhinal cortex (MEC) activity can organize into ultraslow oscillations, with periods of tens of seconds to minutes, and that during those oscillations the cells…
Oscillatory signatures of prediction error in mouse V1: stimulus-dependent, not error-general
José Antonio Velázquez Ruiz
Prior work on the OpenScope Global/Local Oddball dataset reported a gamma-band (~30-90 Hz) power increase in V1 during prediction-error trials, interpreted as a general oscillatory signature of sensory prediction error.
Support S3 read in the AqNWB C++ API
Oliver Rübel, Ben Dichter
The goal of the project is to support data streaming from NWB files in S3 via the AqNWB C++ API using the HDF5 ROS3 driver.

Everything is on campus
Participants stay in private rooms on the Janelia campus, a few minutes' walk from the rooms where sessions run. Meals are provided, the kitchen handles dietary requirements, and the evenings belong to the pub, the pond, and the ping-pong table. For six days, the people who can answer your question are in the building.
The school is free to attend. Lodging and meals are covered by Janelia, travel is reimbursed up to $550 per participant, and additional support can be requested.
What participants said
“I didn't know it was so easy to get started with analysing open datasets!”
“Why wait for the experiments? Use open data now!”
“Getting familiarized with open data is a critical skill for anyone who wants to think through their hypothesis before diving deep into an experimental project, or who needs to replicate results on new data. This workshop is a fantastic way to understand what data is already out there, how to use open science tools, and how to contribute yourself to the open data community.”
Hear when applications open
Announcements about the school, application dates, and new recordings go out on the NWB mailing list a few times a year.









