Vishal Soin

I currently lead the development of modular ADAS Level-2 scenarios using Unreal Engine and CARLA. My work includes procedural scenario generation, multi-sensor integration (LiDAR, RADAR, USS, camera), and Python-based automation to simulate and analyze safety-critical urban interactions like AEB at intersections. Prior to this, I led the development of a Houdini-to-Unreal toolchain at Vayavya Labs, enabling scalable creation of 100+ AI training environments. I also developed batch tools in Python for asset ingestion and implemented USD-based scene configurations, using layering and variants to support multi-LOD environments and sensor permutations. These pipelines accelerated synthetic dataset generation for AI model training and validation in NVIDIA Omniverse.

I’m passionate about architecting pipelines that bridge engineering and content, especially where code and spatial computation meet safety and performance.