VinRobotics - Edge AI & Model Optimization

We optimize and deploy LLMs, ASR, VLM and VLA (Vision-Language-Action) models on real-world systems.

Featured Projects

• vla.cpp Native C++ inference runtime for Vision-Language-Action models, built for low-latency robotic deployment.

• Model Quantization Recipes Practical recipes for quantizing and deploying LLM, ASR, VLM, and VLA models on real-world systems.

What we do

• Optimization: quantization (INT8/INT4/FP8/NVFP4), pruning, distillation, ...
• Deployment: VLLM, TensorRT, ONNX Runtime, edge runtimes
• Systems: real-time pipelines (vision, audio, language, action)

Focus

• Edge devices (Jetson, SoCs)
• Robotics & VLA systems
• Latency, stability, deployability

Philosophy

Optimization = model + runtime + system