It's hard to believe that 2023 marks 20 years of PyCon US. From relatively modest beginnings to serving thousands of in-person attendees—and countless more who participate online and watch talks (sometimes years after they were given)—is a testament to the strength, diversity, and resilience of the Python community.
Since 2011, PyCon US organizers have reserved a row of booths for early-stage technology companies doing new and interesting things with Python. Startup Row has been a fixture of PyCon US for over half the conference's run, and we're pleased to have given over 120 startups the opportunity to share what they're working on with the Python community. Many return as sponsors, and several have gone on to achieve billion-dollar success.
The application process is always competitive, but this year's applicant pool was particularly impressive.
PyCon organizers and the Python Software Foundation are thrilled to welcome seven companies to Startup Row at PyCon US 2023.
The company's CEO, Kanjun Qiu, was Chief of Staff at Dropbox and later led AI recruiting company Sourceress, before starting The Archive, Outset Capital, and eventually Generally Intelligent with Josh Albrecht. Albrecht, who serves as CTO of Generally Intelligent and had worked with Qiu on the aforementioned ventures, was a machine learning-focused academic researcher, helped build Addepar as an early engineer, and was a Thiel Fellow mentor.
"We started Generally Intelligent after our recruiting company, which was a great business, but recruiting felt like an insufficiently ambitious problem to solve," said Albrecht. "We hit upon some fundamental fields in AI research — self-supervised learning, reinforcement learning, etc. — and we've been hard at work for the past couple years on those problems. We're most excited about applying reinforcement learning which act in digital environments: web browsers, desktops, and coding environments."
Neptyne was co-founded by Douwe Osinga, author of the Deep Learning Cookbook for O'Reilly and an artificial life hobbyist, who previously served stints as a Google staff software engineer, CEO and founder of Triposo (acq'd), and senior director of engineering at Sidewalk Labs for its Delve product. Co-founder Jack Amadeo was also at Sidewalk Labs prior to starting Neptyne, where he served as principal engineer of Delve, a generative design tool for city planning.
"Spreadsheets may sound like an orthogonal problem space, but the back-end of Delve produced a pro forma estimate of real estate development costs, which took the form of a spreadsheet. So we've been thinking about computable, extensible spreadsheets for a long time." said Osinga. "Besides, [co-founder Jack Amadeo] and I really love programming. It just made sense to extend the spreadsheet environment with Python." The company is part of the most recent Y Combinator batch.
"Before building a time series data company, we were practitioners," said co-founder Christian Challu. "We started by building the tools we needed for forecasting and anomaly detection. We're always going to be rooted in open source, and we want to keep our packages available to anyone who wants to download and include them in their projects, but we also want to make these tools available via an enterprise-ready API."
Challu continued, "Kind of like what foundation models are doing for NLP, we're building foundation models for time series data."
Co-founders Max Mergenthaler Canseco, Fede Garza Ramírez, Kin Gutiérrez, and Christian Challu are a mix of YC and TechStars founders and AI PhD research scientists aiming to set the standard in statistical, econometric, and neural forecasting algorithms for time series data.
"We started with an observation: It's difficult to get people to use a new tool to get stuff done. I encountered that while working on my PhD, trying to convince researchers to use tools I made [with another platform]," said co-founder Devin Petersohn. "Realizing that Pandas was a widely-used tool, and that datasets are getting bigger and bigger, we built Ponder to make a new scale of data analysis accessible to as many developers as possible."
The company is led by CEO Doris Lee (PhD, Berkeley's RISE Lab), president Aditya Parameswaran (who is also an associate professor at the School of Information at UC Berkeley), and Devin Petersohn, who developed Modin as part of his PhD research at Berkeley's RISELab. Ponder publicly announced $7 million in seed funding back in March 2022; Lightspeed led that deal. Ponder's presence on Startup Row really is one of those "circle of life" situations, considering that the Pandas data science package was itself developed and released by a Startup Row company—financial data company Lambda Foundry, which was in Startup Row's inaugural batch back in 2012.
Thanks to organizers and volunteers who tirelessly work to make PyCon US such a positive, engaging experience, whether that's in-person or virtually. (And, to break the fourth wall for a moment: thanks for giving me the opportunity to help out with a small but very fun part of the conference!)
Thanks also to the dozens of entrepreneurs who took time out of their no-doubt busy schedules to apply to Startup Row. Your time and attention is valuable, and we appreciate all the thought and consideration that went into your applications.
Penultimately, a special thanks goes out to the selection committee which helps score and rank Startup Row applications. This would literally not be possible without the expertise these folks bring to the table, especially since this year was probably the most competitive applicant pool yet.
Finally, thanks to the 7 companies that were able to make it to Salt Lake City and share what they're working on with the over 2,000 folks who come to PyCon US. Thank you, all, again.
Since 2011, PyCon US organizers have reserved a row of booths for early-stage technology companies doing new and interesting things with Python. Startup Row has been a fixture of PyCon US for over half the conference's run, and we're pleased to have given over 120 startups the opportunity to share what they're working on with the Python community. Many return as sponsors, and several have gone on to achieve billion-dollar success.
The application process is always competitive, but this year's applicant pool was particularly impressive.
PyCon organizers and the Python Software Foundation are thrilled to welcome seven companies to Startup Row at PyCon US 2023.
Startup Row at PyCon US 2023
Generally Intelligent
It's 2023 and all debate about whether artificial intelligence has finally gone mainstream is pretty much settled. Generally Intelligent is an independent research company developing general-purpose AI agents with human-like intelligence. The company is founded on a motivation to build the tools its founders and team want to use. Examples include Jupyter Ascending and Avalon, a reinforcement learning simulator that is 100x faster than Minecraft.The company's CEO, Kanjun Qiu, was Chief of Staff at Dropbox and later led AI recruiting company Sourceress, before starting The Archive, Outset Capital, and eventually Generally Intelligent with Josh Albrecht. Albrecht, who serves as CTO of Generally Intelligent and had worked with Qiu on the aforementioned ventures, was a machine learning-focused academic researcher, helped build Addepar as an early engineer, and was a Thiel Fellow mentor.
"We started Generally Intelligent after our recruiting company, which was a great business, but recruiting felt like an insufficiently ambitious problem to solve," said Albrecht. "We hit upon some fundamental fields in AI research — self-supervised learning, reinforcement learning, etc. — and we've been hard at work for the past couple years on those problems. We're most excited about applying reinforcement learning which act in digital environments: web browsers, desktops, and coding environments."
Neptyne
Markdown creator and Daring Fireball blogger John Gruber has said that Excel is the gateway drug to programming. Enter Neptyne, a great web-based collaborative spreadsheet editor with first-class support for Python. Neptyne is great for veteran spreadsheet jockeys, Python newbies who prefer a more visual interface, and data folks looking for a responsive environment in which to prototype and test data enrichment and transformation flows.Neptyne was co-founded by Douwe Osinga, author of the Deep Learning Cookbook for O'Reilly and an artificial life hobbyist, who previously served stints as a Google staff software engineer, CEO and founder of Triposo (acq'd), and senior director of engineering at Sidewalk Labs for its Delve product. Co-founder Jack Amadeo was also at Sidewalk Labs prior to starting Neptyne, where he served as principal engineer of Delve, a generative design tool for city planning.
"Spreadsheets may sound like an orthogonal problem space, but the back-end of Delve produced a pro forma estimate of real estate development costs, which took the form of a spreadsheet. So we've been thinking about computable, extensible spreadsheets for a long time." said Osinga. "Besides, [co-founder Jack Amadeo] and I really love programming. It just made sense to extend the spreadsheet environment with Python." The company is part of the most recent Y Combinator batch.
Nixtla
Nixtla is developing the open source time series ecosystem with roughly 5,200 Github stars across its suite of projects, the most popular of which are StatsForecast and NeuralForecast."Before building a time series data company, we were practitioners," said co-founder Christian Challu. "We started by building the tools we needed for forecasting and anomaly detection. We're always going to be rooted in open source, and we want to keep our packages available to anyone who wants to download and include them in their projects, but we also want to make these tools available via an enterprise-ready API."
Challu continued, "Kind of like what foundation models are doing for NLP, we're building foundation models for time series data."
Co-founders Max Mergenthaler Canseco, Fede Garza Ramírez, Kin Gutiérrez, and Christian Challu are a mix of YC and TechStars founders and AI PhD research scientists aiming to set the standard in statistical, econometric, and neural forecasting algorithms for time series data.
Ponder
Ponder builds on the most popular tools for data science and AI—NumPy and Pandas—extending and scaling these tools to operate directly in enterprise data warehouses like Snowflake, BigQuery, and Redshift, all from an IDE or notebook. Ponder is also the primary maintainer of Modin, an open-source drop-in replacement for Pandas."We started with an observation: It's difficult to get people to use a new tool to get stuff done. I encountered that while working on my PhD, trying to convince researchers to use tools I made [with another platform]," said co-founder Devin Petersohn. "Realizing that Pandas was a widely-used tool, and that datasets are getting bigger and bigger, we built Ponder to make a new scale of data analysis accessible to as many developers as possible."
The company is led by CEO Doris Lee (PhD, Berkeley's RISE Lab), president Aditya Parameswaran (who is also an associate professor at the School of Information at UC Berkeley), and Devin Petersohn, who developed Modin as part of his PhD research at Berkeley's RISELab. Ponder publicly announced $7 million in seed funding back in March 2022; Lightspeed led that deal. Ponder's presence on Startup Row really is one of those "circle of life" situations, considering that the Pandas data science package was itself developed and released by a Startup Row company—financial data company Lambda Foundry, which was in Startup Row's inaugural batch back in 2012.
Predibase
Predibase is a declarative configuration and automation framework for machine learning.
"The problem we wanted to solve was the inefficiency and long time-to-value for most machine learning development," said CEO Piero Molino. "Machine learning workflows are also complex, and often inaccessible to all but the best-resourced companies. Instead of building bespoke models and frameworks for everyone, we [at Predibase] are building a simple, declarative, multi-modal machine learning platform. So, with a simple configuration file, people are able to train and deploy machine learning models without needing to solve scaling and reliability challenges."
The company was founded by the creators of low-code deep learning toolbox Ludwig (8.9k stars on Github) and Horovod (13.2k stars on Github), a distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. Prior to Predibase, CEO Piero Molino was a staff ML/NLP research scientist at Stanford and helped start Uber AI Labs, where he created Ludwig and worked with Predibase CTO Travis Addair, who served as lead maintainer for Horovod.
Pynecone is co-founded by Nikhil Rao, who previously spent over 3 years as an engineer in Apple's Special Projects Group after the tech giant bought autonomous vehicle technology company Drive.ai, where Rao worked as a software engineer. Co-founder Alek Petuskey was previously a machine learning engineer at Ancestry after completing research and software engineering internships at Berkeley SETI, Bender Lab, and NASA Ames. Pynecone is part of the most recent Y Combinator batch.
Rao said, "Current low-code frameworks may be accessible, but they're also limited in their capabilities. We're trying to create a framework for Python developers to build full web applications, without running the risk of outgrowing the framework they started with."
Pynecone is currently hiring, including for its founding full-stack engineer.
Founded by the original creators of the open-source Apache Sedona distributed spatial data processing framework (1.4k Github stars). CEO Mo Sarwat was previously an associate professor of computer science at Arizona State University. Co-founder and engineering lead Jia Yu is on leave from a tenure-track professorship at Washington State University and was a graduate research and teaching assistant at ASU.
Sarwat said, "We're focused on geospatial data, and Python is one of the most popular languages for interacting with data in general, but geospatial data in particular. There's no better place to connect with the Python community than PyCon." Wherobots is currently hiring for infrastructure and developer relations roles.
"The problem we wanted to solve was the inefficiency and long time-to-value for most machine learning development," said CEO Piero Molino. "Machine learning workflows are also complex, and often inaccessible to all but the best-resourced companies. Instead of building bespoke models and frameworks for everyone, we [at Predibase] are building a simple, declarative, multi-modal machine learning platform. So, with a simple configuration file, people are able to train and deploy machine learning models without needing to solve scaling and reliability challenges."
The company was founded by the creators of low-code deep learning toolbox Ludwig (8.9k stars on Github) and Horovod (13.2k stars on Github), a distributed training framework for TensorFlow, Keras, PyTorch, and Apache MXNet. Prior to Predibase, CEO Piero Molino was a staff ML/NLP research scientist at Stanford and helped start Uber AI Labs, where he created Ludwig and worked with Predibase CTO Travis Addair, who served as lead maintainer for Horovod.
Pynecone
Pynecone delivers on the promise of pure Python web applications built on an open source framework. And it has seen some pretty serious community traction—including 8.1k Github stars—in the few short months since they launched in December 2022.Pynecone is co-founded by Nikhil Rao, who previously spent over 3 years as an engineer in Apple's Special Projects Group after the tech giant bought autonomous vehicle technology company Drive.ai, where Rao worked as a software engineer. Co-founder Alek Petuskey was previously a machine learning engineer at Ancestry after completing research and software engineering internships at Berkeley SETI, Bender Lab, and NASA Ames. Pynecone is part of the most recent Y Combinator batch.
Rao said, "Current low-code frameworks may be accessible, but they're also limited in their capabilities. We're trying to create a framework for Python developers to build full web applications, without running the risk of outgrowing the framework they started with."
Pynecone is currently hiring, including for its founding full-stack engineer.
Wherobots
Wherobots is building the database platform for geospatial analytics and AI.Founded by the original creators of the open-source Apache Sedona distributed spatial data processing framework (1.4k Github stars). CEO Mo Sarwat was previously an associate professor of computer science at Arizona State University. Co-founder and engineering lead Jia Yu is on leave from a tenure-track professorship at Washington State University and was a graduate research and teaching assistant at ASU.
Sarwat said, "We're focused on geospatial data, and Python is one of the most popular languages for interacting with data in general, but geospatial data in particular. There's no better place to connect with the Python community than PyCon." Wherobots is currently hiring for infrastructure and developer relations roles.
How You Can Support Startup Row Companies
There are many ways to support Startup Row companies, both at PyCon and afterward:- Visit Startup Row! As the name implies, it's literally a row of booths for startups at PyCon US. You can't miss it!
- Join a Startup Row company! Many companies on Startup Row during the main conference will also be present at the PyCon Jobs Fair on Sunday.
- Contribute to their open source projects! Many of the companies on Startup Row this year have strong roots in open source, and new contributors are always welcome.
- Offer constructive feedback and help! Startups thrive on constructive feedback. Share your perspectives and offer to help out if you think it can benefit the team.
Thank You's and Acknowledgements
Startup Row would not be possible without a lot of trust, logistical support, and hard work from countless folks.Thanks to organizers and volunteers who tirelessly work to make PyCon US such a positive, engaging experience, whether that's in-person or virtually. (And, to break the fourth wall for a moment: thanks for giving me the opportunity to help out with a small but very fun part of the conference!)
Thanks also to the dozens of entrepreneurs who took time out of their no-doubt busy schedules to apply to Startup Row. Your time and attention is valuable, and we appreciate all the thought and consideration that went into your applications.
Penultimately, a special thanks goes out to the selection committee which helps score and rank Startup Row applications. This would literally not be possible without the expertise these folks bring to the table, especially since this year was probably the most competitive applicant pool yet.
Finally, thanks to the 7 companies that were able to make it to Salt Lake City and share what they're working on with the over 2,000 folks who come to PyCon US. Thank you, all, again.