Pau Labarta Bajo
@paulabartabajo.com
Teaching Real World ML at http://realworldml.net
- ML Freelance engineer
- Maths Olympian
- Father of 1.. sorry 2 kids
- ML Freelance engineer
- Maths Olympian
- Father of 1.. sorry 2 kids
Intelligence is neither human nor artificial.
It is just augmented.
It is just augmented.
April 25, 2025 at 9:27 AM
Intelligence is neither human nor artificial.
It is just augmented.
It is just augmented.
Our brain is like a neural network, and learning is like doing gradient descent.
We do something, compare with what we wanted to get, compute the error, reflect and readjust our internal parameters.
We do something, compare with what we wanted to get, compute the error, reflect and readjust our internal parameters.
March 10, 2025 at 11:11 AM
Our brain is like a neural network, and learning is like doing gradient descent.
We do something, compare with what we wanted to get, compute the error, reflect and readjust our internal parameters.
We do something, compare with what we wanted to get, compute the error, reflect and readjust our internal parameters.
Companies to not look for online-course-certificate collectors.
Companies look for problem solvers.
Find a problem you are interested in, build your own solution, put on Gihtub and explain what you solved, and how you solved it in the README file.
Companies look for problem solvers.
Find a problem you are interested in, build your own solution, put on Gihtub and explain what you solved, and how you solved it in the README file.
March 7, 2025 at 11:11 AM
Companies to not look for online-course-certificate collectors.
Companies look for problem solvers.
Find a problem you are interested in, build your own solution, put on Gihtub and explain what you solved, and how you solved it in the README file.
Companies look for problem solvers.
Find a problem you are interested in, build your own solution, put on Gihtub and explain what you solved, and how you solved it in the README file.
ML projects in the real world are very complex.
Training the model in a Jupyter notebook is the easy part.
Complexity and tech debt pop everywhere:
- data ingestion
- deployment
- re-training
- monitoring
MLOps is a set of best practices that make your life easier.
Training the model in a Jupyter notebook is the easy part.
Complexity and tech debt pop everywhere:
- data ingestion
- deployment
- re-training
- monitoring
MLOps is a set of best practices that make your life easier.
February 28, 2025 at 3:19 PM
ML projects in the real world are very complex.
Training the model in a Jupyter notebook is the easy part.
Complexity and tech debt pop everywhere:
- data ingestion
- deployment
- re-training
- monitoring
MLOps is a set of best practices that make your life easier.
Training the model in a Jupyter notebook is the easy part.
Complexity and tech debt pop everywhere:
- data ingestion
- deployment
- re-training
- monitoring
MLOps is a set of best practices that make your life easier.
Behind every great data scientist...
... there is a greater data engineer.
... there is a greater data engineer.
February 26, 2025 at 2:56 PM
Behind every great data scientist...
... there is a greater data engineer.
... there is a greater data engineer.
Kubernetes for ML engineers
www.realworldml.net/blog/kuberne...
www.realworldml.net/blog/kuberne...
Kubernetes for ML engineers
www.realworldml.net
February 23, 2025 at 11:03 PM
Kubernetes for ML engineers
www.realworldml.net/blog/kuberne...
www.realworldml.net/blog/kuberne...
RAG is what happens when your data engineering pipelines work.
Wreck is what happens when they don't.
Wreck is what happens when they don't.
February 22, 2025 at 6:31 AM
RAG is what happens when your data engineering pipelines work.
Wreck is what happens when they don't.
Wreck is what happens when they don't.
Plastic is the worst invention of the 20th century
February 18, 2025 at 5:27 PM
Plastic is the worst invention of the 20th century
Water falls from the sky.
Clean data doesn't.
Thank you my dear Data Engineer.
Clean data doesn't.
Thank you my dear Data Engineer.
February 14, 2025 at 10:44 PM
Water falls from the sky.
Clean data doesn't.
Thank you my dear Data Engineer.
Clean data doesn't.
Thank you my dear Data Engineer.
Your Python code is not working?
A breakpoint() is all you need.
A breakpoint() is all you need.
February 12, 2025 at 2:53 PM
Your Python code is not working?
A breakpoint() is all you need.
A breakpoint() is all you need.
Reposted by Pau Labarta Bajo
Press release by the appropriations subcommittee on NIH budget slashing by @delauro.house.gov 🙏
Trump Team Dismantles Efforts to Find a Cure for Cancer and Other Deadly Disorders and Diseases
President Trump’s Administration is imposing a new policy to cap indirect costs for National Institutes of Health (NIH) research grants at 15 percent, which would cut billions of dollars in funding fo...
democrats-appropriations.house.gov
February 10, 2025 at 4:59 AM
Press release by the appropriations subcommittee on NIH budget slashing by @delauro.house.gov 🙏
We humans are a wonderful product of life evolution.
Not the owners of planet Earth.
Not the owners of planet Earth.
February 9, 2025 at 10:04 PM
We humans are a wonderful product of life evolution.
Not the owners of planet Earth.
Not the owners of planet Earth.
Ever felt like this?
February 6, 2025 at 9:43 AM
Ever felt like this?
A breakpoint() is all you need.
February 5, 2025 at 1:17 PM
A breakpoint() is all you need.
AI AI AI
MLuele tanto!
MLuele tanto!
February 5, 2025 at 10:09 AM
AI AI AI
MLuele tanto!
MLuele tanto!
I remember the good old days where Torchflow was all you needed to know to be called a Deep Learning expert...
... or was it Tensortorch ?
Never mind
... or was it Tensortorch ?
Never mind
February 4, 2025 at 11:22 PM
I remember the good old days where Torchflow was all you needed to know to be called a Deep Learning expert...
... or was it Tensortorch ?
Never mind
... or was it Tensortorch ?
Never mind
You've written a FastAPI app and you need to deploy it to AWS?
Here is a library that will make your life WAAAAAAAY easier
github.com/Kludex/mangum
Here is a library that will make your life WAAAAAAAY easier
github.com/Kludex/mangum
GitHub - Kludex/mangum: AWS Lambda support for ASGI applications
AWS Lambda support for ASGI applications. Contribute to Kludex/mangum development by creating an account on GitHub.
github.com
February 3, 2025 at 8:38 PM
You've written a FastAPI app and you need to deploy it to AWS?
Here is a library that will make your life WAAAAAAAY easier
github.com/Kludex/mangum
Here is a library that will make your life WAAAAAAAY easier
github.com/Kludex/mangum
The data engineer moves the data.
The data scientist talks to the data.
The ML engineer squeezes the data.
The business loves the data.
The data scientist talks to the data.
The ML engineer squeezes the data.
The business loves the data.
February 3, 2025 at 5:26 PM
The data engineer moves the data.
The data scientist talks to the data.
The ML engineer squeezes the data.
The business loves the data.
The data scientist talks to the data.
The ML engineer squeezes the data.
The business loves the data.
$ pip install freetime
February 2, 2025 at 11:23 AM
$ pip install freetime
There is (very) intelligent life beyond LLMs...
February 1, 2025 at 2:28 PM
There is (very) intelligent life beyond LLMs...
Training ML models on static data is just 1 piece of the puzzle.
Real-word ML systems need 2 more pieces:
-> Feature pipeline -> ingests raw data and produce fresh features.
-> Inference pipeline -> generates fresh predictions from the fresh features and your production model.
Real-word ML systems need 2 more pieces:
-> Feature pipeline -> ingests raw data and produce fresh features.
-> Inference pipeline -> generates fresh predictions from the fresh features and your production model.
January 30, 2025 at 8:27 AM
Training ML models on static data is just 1 piece of the puzzle.
Real-word ML systems need 2 more pieces:
-> Feature pipeline -> ingests raw data and produce fresh features.
-> Inference pipeline -> generates fresh predictions from the fresh features and your production model.
Real-word ML systems need 2 more pieces:
-> Feature pipeline -> ingests raw data and produce fresh features.
-> Inference pipeline -> generates fresh predictions from the fresh features and your production model.
You don’t need the latest LLM.
You just need an Excel sheet
You just need an Excel sheet
January 28, 2025 at 10:20 PM
You don’t need the latest LLM.
You just need an Excel sheet
You just need an Excel sheet
US innovates.
China scales.
Europe is on holidays.
China scales.
Europe is on holidays.
January 27, 2025 at 10:57 PM
US innovates.
China scales.
Europe is on holidays.
China scales.
Europe is on holidays.
US innovates.
China scales.
Europe is on holidays*
China scales.
Europe is on holidays*
January 27, 2025 at 9:56 PM
US innovates.
China scales.
Europe is on holidays*
China scales.
Europe is on holidays*
Without data engineering...
... there is no LLM engineering.
... there is no LLM engineering.
January 26, 2025 at 12:52 PM
Without data engineering...
... there is no LLM engineering.
... there is no LLM engineering.