@nixtlainc started last year as a side project
Today we reached 1 million downloadsπ
Our goal is to shake the time series industry and make state-of-the-art algorithms available for everyone π₯
This is how we got there
π§΅1/11
#reinventforecasting
Today we reached 1 million downloadsπ
Our goal is to shake the time series industry and make state-of-the-art algorithms available for everyone π₯
This is how we got there
π§΅1/11
#reinventforecasting
@nixtlainc's ecosystem consists (now) of 5 #python libraries
Focus: speed, scalability, and accuracy π
Some features:
* @scikit_learn syntax
* Native support for #spark, @raydistributed, and @dask_dev
* Models! Eg #arima, #LightGBM, #NeuralNetworks, #transformers π€
2/11
Focus: speed, scalability, and accuracy π
Some features:
* @scikit_learn syntax
* Native support for #spark, @raydistributed, and @dask_dev
* Models! Eg #arima, #LightGBM, #NeuralNetworks, #transformers π€
2/11
We are really proud of the open-source adoption
Repos from Amazon, Mozilla, and DataBricks use us as dependencies πββοΈ
We have contributions from people working at H20, Microsoft, Google, Facebook, SalesForce, Oracle, Shopify, AT&T, Blueyonder, Stanford, MIT and UCL π
3/11
Repos from Amazon, Mozilla, and DataBricks use us as dependencies πββοΈ
We have contributions from people working at H20, Microsoft, Google, Facebook, SalesForce, Oracle, Shopify, AT&T, Blueyonder, Stanford, MIT and UCL π
3/11
@nixtlainc is used in production for
π Anomaly detection at a big ride-sharing app
π Hierarchical Reconciliation at one of the top consultancy firms in the world
π Forecasting in big names from finance and energy
(We can't reveal names now. But no, not crypto) π³
4/11
π Anomaly detection at a big ride-sharing app
π Hierarchical Reconciliation at one of the top consultancy firms in the world
π Forecasting in big names from finance and energy
(We can't reveal names now. But no, not crypto) π³
4/11
Reproducibility
We claimed we were faster, more accurate, and really scalable, and we backed it with reproducible experiments π₯Όβ²οΈπ¬
5/11
We claimed we were faster, more accurate, and really scalable, and we backed it with reproducible experiments π₯Όβ²οΈπ¬
5/11
Extended Family π©βπ©βπ§βπ¦
We partnered with top open-source companies. π€
Thanks to: @openbb_finance, @fugue_project, @MindsDB, @LightningAI, and @raydistributed.
More exciting things are yet to come.
We are always happy about new collaborations.
Looking at you! π₯
6/11
We partnered with top open-source companies. π€
Thanks to: @openbb_finance, @fugue_project, @MindsDB, @LightningAI, and @raydistributed.
More exciting things are yet to come.
We are always happy about new collaborations.
Looking at you! π₯
6/11
We believe in first principles and science.
Our favorite publication this year:
N-HITS: significant contribution to explainable deep learning for time series. (Accepted at AAAI π)
Coauthoring with @BOreshkin has been an honor and privilege for our research team. π€
7/11
Our favorite publication this year:
N-HITS: significant contribution to explainable deep learning for time series. (Accepted at AAAI π)
Coauthoring with @BOreshkin has been an honor and privilege for our research team. π€
7/11
With a little help from my friends πΆ
Thanks to our wonderful advisors: @seanjtaylor, @kolencherry, @adamberke, @DevangSachdev
(Yes, we also thank our GREAT investors. Announcement coming soon.)
8/11
Thanks to our wonderful advisors: @seanjtaylor, @kolencherry, @adamberke, @DevangSachdev
(Yes, we also thank our GREAT investors. Announcement coming soon.)
8/11
Shoulders of giants. π³
We are very humbled that old heroes of ours showed support and love.
Special thanks to @robjhyndman (THE guy on the time series field)
And also special thanks to @ClementDelangue, @_willfalcon, and @seanjtaylor for believing in us
9/11
We are very humbled that old heroes of ours showed support and love.
Special thanks to @robjhyndman (THE guy on the time series field)
And also special thanks to @ClementDelangue, @_willfalcon, and @seanjtaylor for believing in us
9/11
This is a biased selection of what really happened last year #selectionbias
Downlights π€
We got rejected by more than 50 VCs
We did not get into YC or DG
ICML passed on our paper
And... we forgot to submit on time for NeurIPS
That made us insecure and sad at the momentπ’
10/11
Downlights π€
We got rejected by more than 50 VCs
We did not get into YC or DG
ICML passed on our paper
And... we forgot to submit on time for NeurIPS
That made us insecure and sad at the momentπ’
10/11
Coming next...
π ...week: exciting new features and a big surprise on Wednesday. #reinventforecasting
π ...month: exciting content with @databricks and @anyscalecompute.
π§ ...year: Transfer-learning and Large Time Series Models for the community.
Stay tuned!
11/11
π ...week: exciting new features and a big surprise on Wednesday. #reinventforecasting
π ...month: exciting content with @databricks and @anyscalecompute.
π§ ...year: Transfer-learning and Large Time Series Models for the community.
Stay tuned!
11/11
While writing this tweet, we also forgot to replace DG with @danielgross and @aigrant_.
β’ β’ β’
Missing some Tweet in this thread? You can try to
force a refresh
γ
