Computer Vision Expert

Key Responsibilities

• Architect and implement computer vision models for

  object detection, tracking, face recognition, and re-identification

  (person tracking across cameras).
• Lead

  model optimization

  for real-time inference (ONNX, TensorRT, mixed precision, pruning/quantization).
• Develop data pipelines for

  multi-camera ingestion, synchronization, and annotation

  .
• Experience in

  SOTA architectures

  (e.g., YOLOv8, SAM, CLIP, RT-DETR, DeepSORT, etc.).
• Work on

  multi-GPU and distributed inference systems

  , including CUDA optimization and batching.
• Deploy models using

  FastAPI / gRPC microservices

  and monitor system performance.
• Mentor data scientists and review their experimental results.
• Collaborate with backend engineers to integrate computer vision APIs into larger products.

Required Skills

• Languages:

  Python (advanced), C++ (preferred for performance modules).

• Frameworks:

  PyTorch, TensorFlow, ONNX Runtime, OpenCV, TorchServe, or Triton.

• Tools:

  Docker, CUDA, TensorRT, Git, DVC, MLflow.

• Deep Knowledge of:

• Object detection & tracking pipelines (YOLO, DETR, SORT, DeepSORT).
• Face recognition pipelines (ArcFace, InsightFace).
• Multi-threaded and GPU-based inference systems.
• CCTV AI-Based Monitoring and Deployments at scale.

• Deployment:

  Kubernetes-Docker, Zookeeper-kafka, gRPC, FastAPI, or WebSocket-based inference endpoints.

• Bonus:

  Experience with

  multi-camera calibration

  ,

  temporal analytics

  , or

  behavioral analytics

  .

Qualifications

• M.Tech/MS/PhD in Computer Science, Computer Vision, or Applied Machine Learning preferred.
• Bachelors in CS/EE/Maths or any other relevant education with must have exceptional hands-on experience also considered.

Good-to-Have Skills

• Proven ability to lead a team with scientific rigor and hands-on coding.
• Strong documentation and experiment tracking practices.
• Curious, research-driven, and comfortable working in high-performance, fast-iteration environments.