I’m an experienced researcher, developer, and manager working on multimodal language models, reinforcement learning and autonomous navigation.

Language models, RL and Self-driving

Currently, I’m a Principal Applied Scientist at Wayve working on end-to-end self-driving systems.

Areas I’m currently investigating:

• Video LLMs for autonomous driving explainability
• Language models for reasoning in self-driving
• Multiagent RL (small blogpost)
• Model-Based RL, MCTS, and adaptive planning
• Inverse and Offline RL
• Novel View-point Synthesis (like NERFs) and self-driving evaluation

I’m also involved in technical leadership and a bit of management.

Scalable Robotics and Management

Previously, I held the title of VP of Research and Development at BrainCorp. I built a research team from the ground up, scaled up AI, robotics, and machine learning systems from zero to thousands of robots deployed across the world (longer story).

We had a chance to work on:

• Motion planning, optimization, and search algorithms (paper)
• SLAM (pose graphs, particle filters, etc.)
• Perception (DL, computer vision, depth, and lidars - talk)
• Large-scale data analysis and ML

Ages ago, I was part of a Robotic Soccer team and went to a few international championships.

Real and Artificial Brains

Before pivoting into robotics, BrainCorp was a neuroscience company where I worked as a Senior Research Scientist. I worked on simulated human and animal brains and other fun stuff:

• Spiking neurons, brain modeling
• RL, Supervised and Unsupervised learning in biologically-plausible neurons
• Neuromorphic hardware and energy-efficient computation

Ah, also I’m a Computational Neuroscience Ph.D. from a robotic department of MPEI. I’m an author of 40+ patents (just search here for Oleg Sinyavskiy).

A lot of coding

I have been a coder since I can remember myself, so I qualify to make a coding blog :)

Several times in my career, I created company-wide design and coding guidelines. I started in C++ working on city-scale license plate detection systems and telepresence robots. Then I switched to Python and designed software that powers tens of thousands of systems across the world.

I’m a programming language enthusiast and always happy to learn about a new language. I enjoyed picking up Haskell, Go, Rust, Scala, and few others.

License

All the website, the code it has, carries a Creative Commons Attribution 4.0 International License. You can use the content and then sell it (please credit me). All images are created by me using my image editing tools and outputs of Stable Diffusion models under Open RAIL++-M licence and hence also available for commercial use.