Carlo Lepelaars

Image by Ishant Mishra |

Towards A More Profitable S&P500 With Portfolio Optimization

Just published my first blog post on Medium! This new blog post discusses how you can optimize the S&P500 and create portfolios that are more performant compared to the classic S&P500 weighted by market capitalization. As an introduction to portfolio optimization you’ll learn about the equal weighted S&P500, minimum variance portfolio, eigenportfolios and random matrix…
Read more

Crowdcent, logo, investing, NLP, machine learning, numerai, crowdsourcing

open-sourcing numerblox

Happy to announce we are open-sourcing a library at CrowdCent called numerblox, which we have been building and using internally since January. We built it mainly to simplify and scale our weekly inference pipelines, but are also using components of it for training Numerai models. The vision behind numerblox is that while the Numerai community…
Read more

Crowdcent, logo, investing, NLP, machine learning, numerai, crowdsourcing

Joining CrowdCent full-time

Very happy to announce that, starting in January, I’m joining CrowdCent full-time! I will focus on financial data, NLP and Numerai models. CrowdCent’s vision is to decentralize investment management by aligning incentives, crowdsourcing ideas and processing them with machine learning. For more information:

signals, numerai, introduction, nmr

Numerai Signals Blog post

I just released a new blog post on getting started with Numerai Signals using and a sentiment analysis model (FinBERT). Blog post link here Accompanying code link here This post was written for Weights and Biases. More blog posts that I wrote for this platform can be found here.

Transformer, Attention is all you need

Transformer Deep Dive blog post

“If you want to learn something well, explain it.” – Richard Feynman Over the Christmas break I wrote a deep dive about the fundamental building block of so many awesome AI models nowadays (GPT-3, DALL-E, etc.): Transformers! Includes background, PyTorch code and formulas. Hope you learn as much from it as I did writing it!…
Read more

GradCAM, Artificial Intelligence, AI, XAI, Elephant, Keras, Tensorflow

Discussing Explainable AI with Jönköping University professor Beril Sirmacek

The past weeks I have had the honor to talk about Explainable AI (XAI) with Beril Sirmacek for a course she teaches at Jönköping University in Sweden. We talk about how I started with data science and especially computer vision. We go into several XAI techniques. In a follow-up video we also do a deep…
Read more

Convolutional neural networks, leaves, apple trees, plant disease, computer vision, kaggle, weights and biases, wandb

The evolution of mobile CNN architectures

For Weights & Biases I just released a new report researching the evolution of mobile Convolutional Neural Network (CNN) architectures (MobileNet, SqueezeNet, GhostNet, etc.). It also includes a Kaggle Notebook so you compare and experiment with the CNN architectures yourself. Weights & Biases Report Kaggle Notebook

numerai, interview, office hours with arbitrage

Numerai Interview on “Office Hours with Arbitrage”

For the last half year I’ve been active on Numerai, an online Data Science tournament where you build predictive models for stock market trading. When I started with the tournament it felt tough to get started and setup a good model evaluation scheme. As a result of this initial struggle I decided to feature Numerai…
Read more

Weights And Biases, logo

Joining Weights & Biases’ Author program

For the last two weeks I started writing for Weights & Biases (W&B) as part of their author program. They are making very cool tools for machine learning engineers. I will highlight some essential features in upcoming blog posts. My first blog post is on getting started with Numerai, which is a data science tournament…
Read more

stuffed elephant, computer vision, cnns, 3d, rotation

Estimating 3D rotation with deep learning

For the past two months my colleagues and I at VBTI have been working on estimating 3D rotation with deep learning for various applications. Here we show an example demonstration using a laptop webcam and a stuffed elephant (Ollie). 😁