Hebbian Robotics
APIs for robotics data teams to verify data quality for model training
About Hebbian Robotics
Hebbian Robotics builds APIs for robotics data teams to verify data quality for model training. Our customers use us to build quality control pipelines in just a few lines of code, with custom models 10x faster and 8x cheaper than general foundation models. https://hebbianrobotics.com/ https://github.com/Hebbian-Robotics/hflow
Public traction evidence
Each signal links to the public source used for attribution.
- X
Today we're launching Hebbian Robotics (YC S26) @hbr_pbc .
Today we're launching Hebbian Robotics (YC S26) @hbr_pbc . We build APIs for evaluating data quality, without training a robotics model. Data vendors use our API to search and analyze Physical AI data at scale, getting on-demand quality signals and metrics without managing
- X
Dyna’s article on training Dyna-2 with more than 1 million hours of egocentric video is a gold mine for anyone processing multimodal robotics data at scale.
Dyna’s article on training Dyna-2 with more than 1 million hours of egocentric video is a gold mine for anyone processing multimodal robotics data at scale. We built HFlow: an open source implementation of the data infrastructure described in the article. - Airflow for
- X
If you work with large multimodal robotics datasets, then you need APIs to search and analyze them.
If you work with large multimodal robotics datasets, then you need APIs to search and analyze them. We are currently working with data vendors and providers to get signals and metrics on the quality of their data. Check out our launch and reach out if this resonates! @hbr_pbc
- X
We've been speaking with teams who deeply care about data quality, and we noticed that every data team eventually builds similar pipelines for quality checks (QC).
We've been speaking with teams who deeply care about data quality, and we noticed that every data team eventually builds similar pipelines for quality checks (QC). Teams collecting data want to own their quality checks (camera blackout, choppy joint states, occluded hands),
- X
If you had the chance to attend this, you now have an edge in starting your own robotics startup A slight edge, but it can compound, and you should start soon
If you had the chance to attend this, you now have an edge in starting your own robotics startup A slight edge, but it can compound, and you should start soon
- X
We used HFlow to evaluate the latest open weights VLMs for processing egocentric data.
We used HFlow to evaluate the latest open weights VLMs for processing egocentric data. This was based on @buildpbc's Egocentric-10k evaluation, which used Gemini 2.5 Flash to measure hand visibility and active manipulation. How much each model agreed with the reported results:
- X
This release effectively doubles the hours of open source egocentric data on @huggingface This was previously dominated by @buildpbc's Egocentric-100K, which was single monocular camera only That's 25 TB vs 53 TB (Ego-100k vs EgoSuite-100k) as a size comparison Ego-100k also
This release effectively doubles the hours of open source egocentric data on @huggingface This was previously dominated by @buildpbc's Egocentric-100K, which was single monocular camera only That's 25 TB vs 53 TB (Ego-100k vs EgoSuite-100k) as a size comparison Ego-100k also
- X
The most insidious bugs in Physical AI aren't in your model.
The most insidious bugs in Physical AI aren't in your model. They're in your data pipeline. Training data in Physical AI is rapidly growing in both size and complexity. It is no longer just time-synchronized video streams. Training payloads now carry hand positions, depth, and...