Hey there, I'm Daniel Jonas,But most people call me Daniel 😊
a full-blooded musician 
jodaas
SoundCloud Artist
Music Producer from Hamburg, Germany
and master student in the fields of AI and Data Science as well as a student AI & Data Scinentist at Statista+.
jodaas
SoundCloud Artist
About Me
I like AI and music, obviously. Currently, I'm focus on audio generation - exploring diffusion and flow models, LLMs, and hybrid approaches like Diffusion Transformers (DiTs) for SOTA music generation.
I'm particularly interested in neural audio codecs and efficient tokenization strategies for representing audio in latent space using VAEs, GANs, and Vector Quantization for music for downstream generative models. As AI is evolving rapidly across generative models, reinforcement learning, and agentic trends, I continuously explore new research to stay current.
Latest Posts
All posts
Flow and Diffusion Models Part 2: Training Target
A mathematical guide to understanding the training targets of flow-based models and diffusion processes in deep learning.
The Mathematics Behind Flow and Diffusion Models
A deep dive into the mathematical foundations of flow-based models and diffusion processes in deep learning.
My first Blog Post!
An introduction into Neural Networks, Gradient Descent and Backpropagation.
Dual-Midi Transformer
A recap of my work on the Dual-Midi Transformer project, which aims to generate music from MIDI data.
An introduction into Artificial Music Generation
A comprehensive overview of the current state of music generation techniques
Recent Projects
All projectsDual MIDI Transformer
A MIDI Event based transformer architecture, utilizing two seperate transformers operating on the global MIDI events and local event parameters.
Basic MIDI Transformer
A basic transformer architecture for MIDI generation, utilizing a single decoder to enable faster training and inference speed.
Want to collaborate?
Always open to talk, feel free to leave me a message!