Computer Vision Engineer

Role Overview:-

You will work on next-generation Smartbox R&D initiatives focused on cameras, sensors, object detection, access automation, and intelligent monitoring. Your contributions will enhance the reliability, safety, user experience, and automation of our Smartbox locker products.

Roles and Responsibilities:-

Computer Vision Module Development

• Build computer-vision models to detect key user actions such as door opening, door closing, and object placement.

• Develop detection capabilities for misuse or abnormal scenarios to enhance system safety and reliability

Object Detection & Tracking

• Develop object detection models using YOLO, Detectron2, or similar deep-learning frameworks.

• Build and optimize object tracking pipelines for real-time Smartbox operations.

Real-Time Video Stream Processing

• Handle real-time video streams from IP cameras.

• Work with embedded camera inputs.

• Manage and process video feeds using RTSP, USB, or similar interfaces.

Dataset Creation & Management

• Create datasets using footage collected from Smartbox terminals.

• Build datasets from test rigs for model training and validation.

Computer Vision Model Development

• Train computer vision models for high accuracy and reliability.

• Test model performance across various scenarios and datasets.

• Optimize models to reduce latency and improve real-time responsiveness.

Event Detection & Monitoring

• Work on detecting key events such as parcel presence and locker occupancy.

• Develop safety-related monitoring features to identify abnormal or unsafe conditions.

Computer Vision Integration

• Integrate computer vision logic with the Smartbox backend.

• Work with Java-based microservices to ensure smooth data flow and system interoperability.

Edge Deployment

• Deploy computer vision models on edge devices such as Jetson and NUC.

• Work with embedded boards to enable real-time, on-device inference.

Prototype Development

• Build quick prototypes to validate new Smartbox hardware features.

• Test and refine concepts to support rapid R&D experimentation.

Desired Candidate Profile

• Demonstrated ability to build, test, and optimize computer-vision pipelines using Python and OpenCV.
• Hands-on experience with YOLO (v7/v8/v11) or Detectron2 for object detection, model training, and fine-tuning.
• Proven experience in creating, annotating, and training models on custom datasets for real-world applications.
• Strong understanding of RTSP, FFmpeg, and GStreamer for handling real-time camera feeds and video pipelines.
• Knowledge of tracking algorithms such as DeepSort, ByteTrack, or similar frameworks for real-time MOT applications.
• Comfortable working with physical devices, sensors, cameras, and lab environments for model validation.
• Practical experience deploying ML/CV models on edge hardware such as NVIDIA Jetson, mini-PCs, or embedded systems.