

Boris Denisenko
Senior Software Engineer (Android Automotive OS)

Hi, I'm Boris. I’ve been passionate about software engineering for over 17 years. Throughout my career, my path has taken me across diverse domains: from building widely used mobile applications to working closely with embedded systems and advanced automotive technologies. Today, I'm based in Munich, working as a Technical Lead and Senior Software Engineer at Mapbox, where I architect complex navigation solutions for leading global car manufacturers.
My professional journey is driven by the challenge of bridging the gap between raw hardware signals and seamless user experiences. A significant milestone for me was my time working for Jaguar Land Rover project, where I served as the Lead Mobile Developer for their vehicle remote control platform. I also spent time pushing the boundaries of mobile tech as a core engineer on the Mapbox Vision SDK, developing real-time computer vision systems for edge computing—a project recognized for its technical excellence by the Qualcomm Developer Network. While my past projects heavily involved machine learning and computer vision, my core focus today lies in conquering the architectural challenges of Android Automotive OS (AAOS).
My technical foundation is rooted in a Master’s degree in Applied Mathematics and Informatics. In my day-to-day practice, I leverage Kotlin, C++, and robust Android architectures to build high-performance, hardware-integrated systems. Ultimately, I believe that truly great automotive products are born when we translate noisy, complex vehicle data into a reliable and inspiring experience for the person behind the wheel.
Architecting a Real-Time LLM Co-Pilot for Android Automotive: Lessons from MapGPT
In-car voice assistants have historically been rigid and slow. What happens when you integrate real-time vehicle telemetry, dynamic route data, and a streaming LLM directly into an Android navigation app?
This session explores the Android architecture behind MapGPT, a production in-car AI assistant. Moving beyond standard REST APIs, we will dive into managing an always-on WebSocket stream using Kotlin Coroutines and StateFlow to process natural language and seamlessly trigger native UI and vehicle actions.
You will learn how to:
- Design Coroutine/Flow pipelines for streaming LLM responses without UI jank.
- Implement a decoupled Kotlin Middleware pattern to execute actions like routing and climate control.
- Manage complex app state (telemetry, ETA, EV data) to feed the LLM continuous context.
- Build a deterministic replay system to test streaming interactions in CI without a live API or physical car.
