Ángel Igareta

Data Scientist at Klarna⚡️

I'd like to help create the technology of the future. not only use it.

[Ángel Igareta @angeligareta]

Happy to share that I've been promoted to Senior Data Scientist at Klarna! Excited to continue contributing to and exploring new opportunities in this dynamic industry https://t.co/WiNn8qWAAF

[Ángel Igareta @angeligareta]

Are you passionate about #NaturalLanguageProcessing? If so, you're going to love a library I recently encountered that allows you to augment text data! 🔥

For those unfamiliar, data augmentation techniques are used to generate additional synthetic data u…https://t.co/oVonK31OXw

[Ángel Igareta @angeligareta]

Let's say you are done running and tweaking your modeling pipeline and it's time to deploy the initial version of your model, possibly for a demo to stakeholders. 🎉

This process usually takes a great deal of time, perhaps due to a lack of experience in…https://t.co/T7vGDWmiYA

[Ángel Igareta @angeligareta]

💡 Interesting Geek Fact 💡 Surely you know that to initialize an empty list in Python you can either type [] or list(), but did you know that one of these two methods is faster than the other? 🤯

Analyzing the bytecode of these two lines using the built…https://t.co/ysORFDVHef

[Ángel Igareta @angeligareta]

Hyperparameter tuning, essential every time you train a new model to find the best performing configuration for your target model and dataset. You test as many combinations as time allows and get a set of parameters. But to what extent are you sure you ar…https://t.co/lnpmJAC2BI

[Ángel Igareta @angeligareta]

Progress. Not only important to experience it in life, but also when it comes to #programming 😜 Here's a super useful library that every #datascientist should know about! 💡

#TQDM is a Python tool to smoothly see a smart progress meter on your loops. Fo…https://t.co/p3uRvlrchn

[Ángel Igareta @angeligareta]

Exploratory Data Analysis (EDA) can be a challenging and time-consuming task, especially when analyzing a dataset with hundreds of features. However, understanding the various aspects of the data is essential before developing any deep learning model ⚠️. …https://t.co/64gd4HepFt

[Ángel Igareta @angeligareta]

As soon as I came across this paper I wanted to share it with all of you because of its impressive results: 𝐓𝐨𝐰𝐚𝐫𝐝𝐬 𝐑𝐞𝐚𝐥-𝐖𝐨𝐫𝐥𝐝 𝐁𝐥𝐢𝐧𝐝 𝐅𝐚𝐜𝐞 𝐑𝐞𝐬𝐭𝐨𝐫𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐅𝐚𝐜𝐢𝐚𝐥 𝐏𝐫𝐢𝐨𝐫 💡

Instead of…https://t.co/EShN9sYEcw

[Ángel Igareta @angeligareta]

When I developed my master thesis applying CNNs, I extensively used #TensorBoard to analyze the results and get better interpretability of my model. 

I also found that it is possible to project the low-dimensional #embeddings generated by my CNN (output…https://t.co/yJq9hlsX7r

[Ángel Igareta @angeligareta]

After training and hyperparameter tuning of your model, the next step in #DataScience projects is usually to present the final results and make the best model obtained accessible. 📊

One of the most popular ways to do this is through an API. When it come…https://t.co/eQdYFtFzv8

[Ángel Igareta @angeligareta]

After training and hyperparameter tuning of your model, the next step in #DataScience projects is usually to present the final results and make the best model obtained accessible. 📊

One of the most popular ways to do this is through an API. When it come…https://t.co/r2olJQwh1A

[Ángel Igareta @angeligareta]

👇🔥

[Towards Data Science @TDataScience]

"One of the first decisions to be made when training deep neural networks is to select the epoch in which to stop. And it is not an easy one." From @angeligareta, a guide to the "early stopping" approach. https://t.co/65Kus18O6y

[Ángel Igareta @angeligareta]

One of the tools that has been most helpful to me in finding articles for research is Connected Papers, a website that provides an excellent visual approach to discovering similar articles. 

I have been using this software during…https://t.co/LD7DBHWRXd https://t.co/HcX1OOHqNj

[Ángel Igareta @angeligareta]

Now, here's a creative app powered by #Tensorflow! How to remove people in real-time even from complex backgrounds 💡.

But how does it actually work? First, the system detects pixels in a video stream that belong to a human body using the TensorFlow.js B…https://t.co/HqwgYJiUfW

[Ángel Igareta @angeligareta]

Remember the infamous #HarryPotter magic portraits where it was possible to interact with the people in it? What if I told you that there is a cloud service that comes closer to bring your black and white photos back to "life"?

The company MyHeritage int…https://t.co/gEvirtDOcz

[Ángel Igareta @angeligareta]

Do you also think that video conferencing platforms still have enormous room for improvement? Having used them more frequently than ever, you would probably agree that these applications are still a long way from making calls feel…https://t.co/sG4vyZa6qT https://t.co/oTIr5fgKoD

[Ángel Igareta @angeligareta]

This year Google is bringing around 100 AI-powered improvements for #GoogleMaps 🎉

Some of them have really caught my attention, as an example of innovation and application development with a focus on sustainability. Check them out here: https://t.co/M3P4PNkIH6

[Ángel Igareta @angeligareta]

@RieraULL @CristianAbrante @_carlosdg_ Sin duda! Nos hizo mucha ilusión escuchar la noticia de Daute. Espero que todo esté yendo bien! 😄

[Ángel Igareta @angeligareta]

@_carlosdg_ @CristianAbrante It would certainly not have been possible with all of us 💪

[Ángel Igareta @angeligareta]

Woohoo! My team, Gofio's Mafia, came in 29th place overall in the #HashCode Online Qualifications in Spain. Our final score was 9,497,876. 🏆

@_carlosdg_ @CristianAbrante DauteRodriguez