AI & ROBOTICS
Learning how
machines learn.
I'm Nick Sokolov, an ML research engineer working on multimodal learning and embodied AI.
I focus on vision‑language‑action models and robotic manipulation, and investigate how AI models perceive, reason, and represent the world.
FIG. 01 — PERCEPTION TO ACTION
AREAS OF INTEREST
- Vision-Language-Action Models (VLA)
- World Action Models (WAM)
- AI Cognition & Interpretability
- Vision-Language Models (VLM)
- Robot Learning
THE RESEARCH NOTEBOOK
Selected explorations
VISION & LANGUAGE01 /Building image-captioning models, from CNN–LSTM pipelines to attention and Transformer decoders.
Explore the notebookFrom Pixels to Words
How machines learned to describe the world.
COMPUTER VISION02 /Implementing a Vision Transformer step by step, training on CIFAR-10, and inspecting where the model attends.
Explore the notebookVision Transformer, from scratch
From image patches to attention maps.