BREAKING  Creator of Dask now runs a venture-backed company NASA runs his code  //  so does the US Air Force $26M raised to make Python scale Berkeley physics → Chicago PhD → Coiled CEO He walked across Spain and benchmarked dataframes — both made the blog BREAKING  Creator of Dask now runs a venture-backed company NASA runs his code  //  so does the US Air Force $26M raised to make Python scale Berkeley physics → Chicago PhD → Coiled CEO He walked across Spain and benchmarked dataframes — both made the blog
Open Source · Cloud · Python at Scale

Matthew
Rocklin

He wrote the library that lets a laptop's worth of Python push around a thousand cloud machines. Then he built a company so you wouldn't have to think about it.

Creator of Dask CEO, Coiled ex-NVIDIA ex-Anaconda
Matthew Rocklin
Matthew Rocklin — the quiet plumbing behind big Python.
Share LinkedIn Twitter / X Facebook Instagram
The Dispatch

A data scientist opens a Jupyter notebook in New York, types a few lines of ordinary Python, and a thousand machines spin up across a cloud region to chew through a terabyte. She never sees the machines. She never writes a line of Kubernetes. That invisible handoff - from one laptop to a fleet and back - is the thing Matthew Rocklin has spent a decade making feel boring.

Boring is the compliment. The hardest infrastructure is the kind nobody notices. Rocklin created Dask, the open-source library that scales NumPy and Pandas across many cores and many machines, and he co-founded Coiled, the company that wraps that power in a developer experience pleasant enough that you forget it's hard.

PythonDaskNumPyPandas AWSAzureGCPGPUs Distributed SystemsSymPytoolz
By The Numbers
$26M
Total raised by Coiled
2014
Year Dask was born
3
Clouds supported - AWS, Azure, GCP
1000s
Data teams running on it
How He Got Here

From a weekend hack to the engine room of scientific Python

Dask did not arrive as a grand plan. Rocklin set out to parallelize NumPy and Pandas - to let two of Python's most-loved libraries spread their work across more than one core. The fix worked. Then something more interesting happened.

The internals he built to schedule that parallel work turned out to be useful for far more than arrays and dataframes. So he pivoted, exposing those internals as a general-purpose parallel system. Dask stopped being a NumPy accelerator and became a way to scale almost any Python computation - custom code, machine learning, geospatial pipelines, batch jobs.

The career reads like a tour of the places where scientific Python actually gets built. Years at Anaconda (then Continuum Analytics), where Dask first took shape. A stint at NVIDIA, building out the Dask team behind RAPIDS to push the Python data stack onto GPUs. Then, in 2020, the leap into founding Coiled.

It is the classic open-source-to-startup arc, run in the right order: ship the library first, build the company second. The users were already there. The job was to make the hard part easy enough to pay for.

"Increase accessibility to computation, helping us accelerate science and inform policy decisions for the broader public good."

- Matthew Rocklin, on his stated mission
The Timeline

A decade of teaching Python to scale

2014

Creates Dask at Continuum Analytics to parallelize NumPy and Pandas. The side effect becomes the main event.

2018

Joins NVIDIA, building the Dask team behind RAPIDS and bringing GPU acceleration to the Python data world.

2020

Co-founds Coiled Computing and steps in as CEO. Mission: make Dask effortless for large organizations on the cloud.

2021

Coiled raises a $21M Series A. Rocklin keynotes the Dask Distributed Summit.

2024

Coiled publishes public TPC-H benchmarks pitting Spark, Dask, DuckDB and Polars against each other at scale.

The schooling

Physics and mathematics at UC Berkeley. A PhD in computer science from the University of Chicago. He came to distributed systems through equations, not server racks - which may be why he keeps trying to make the racks disappear.

Where The Code Runs

You have probably never heard his name. You have used his work.

NASA & the USAF

Public-sector teams lean on Dask and Coiled for planetary-scale data and analysis - the kind of workloads that do not fit on one machine.

Capital One & Anthem

Finance and health organizations use it for credit risk analysis and large-scale data processing where Python ergonomics matter.

The PyData crowd

Beyond the marquee names, thousands of data scientists and engineers reach for Dask whenever a single laptop stops being enough.

Off The Clock

The blog tells on him

Most founders curate. Rocklin documents. His personal site, matthewrocklin.com, runs under the unglamorous title "Working Notes," and the posts wander from open-source maintenance and startup practice to workplace dynamics and the parts of the job that wear people down.

Tucked among the engineering writeups: an account of walking the Camino across Spain. The same temperament that benchmarks four dataframe engines in public and publishes the receipts also walks five hundred miles and writes about that, too.

He runs a New York-headquartered company from Austin, Texas, which is its own small comment on the remote-first world his tools helped enable. And he is on Mastodon rather than chasing every platform - a quietly opinionated choice from someone who has spent a career thinking about open systems.

When he demoed Coiled to the geospatial crowd at SatCamp, what won them over was not raw speed. It was how simple the architecture was. For a man who builds complicated machinery, the recurring goal is to hide it.

Field Notes

Five things that fit on a postcard

  • His GitHub handle, mrocklin, sits behind some of the most-used plumbing in the Python data world.
  • Dask began as an effort to speed up Pandas. It now schedules terabyte-scale jobs.
  • He writes about open-source burnout as readily as he writes about code.
  • He contributed to SymPy, toolz, and Theano before Dask made his name.
  • He benchmarks his own product against rivals in public - and shows the numbers.

"The aim is simple and stubborn: make large-scale computation as accessible as writing ordinary Python - so scientists and analysts can solve bigger problems without becoming distributed-systems experts."

- The throughline of everything he builds