We are developing a large-scale video monitoring and intelligent video analytics system for a government project.
Responsibilities
- Develop production systems for video analytics based on Computer Vision
- Process dozens and hundreds of RTSP video streams with minimal latency
- Build and optimize pipelines using NVIDIA DeepStream, GStreamer, and TensorRT
- Develop and implement Computer Vision models (YOLO, ViT, Video Transformers, VLM)
- Optimize inference on GPU/CPU (TensorRT, OpenVINO, ONNX, Quantization)
- Integrate Computer Vision with LLM and multimodal models for generating intelligent analytics from video data
- Develop AI services using FastAPI/gRPC, Docker, Kubernetes
- Build RAG and Agentic AI solutions (LangGraph, vLLM, Function Calling)
- Design the architecture of AI components and participate in technical decision-making
Requirements
- Commercial experience in Computer Vision and Video Analytics
- Practical experience with RTSP, DeepStream, GStreamer, TensorRT
- Excellent knowledge of PyTorch, OpenCV, YOLO
- Experience in model optimization (ONNX, TensorRT, OpenVINO, CUDA)
- Understanding of modern Vision Transformers (ViT, DETR, Swin, TimeSformer, ViViT) and multimodal models (LLaVA, Qwen-VL)
- Experience with LLM, RAG, and agentic systems is a plus
- Experience in developing production AI systems and ML services
Optional
- DeepStream SDK
- CUDA
- Triton Inference Server
- Kafka
- Kubernetes
- LangGraph
- vLLM
- MLflow
