Computer Vision & AI Intern

About the Role

We are building an on-premise computer vision system to detect operator fatigue and man-down events on our production floor. This system uses existing cameras, with alerts sent to the operator first, escalating only for genuine safety events. No video leaves the machine. This project focuses on safety and productivity monitoring to prevent fatigue-related incidents.

What You'll Build

• Ingest:
• Audit existing cameras for RTSP access, frame rate, night IR coverage, and pixels-per-meter.
• Build a multi-stream pipeline with hardware decode.
• Detection:
• Implement person detection and tracking.
• Perform pose estimation for posture (standing, slumped, prone).
• Calculate motion-energy from keypoint variance.
• Detect eye-state/PERCLOS where resolution permits.
• Fuse detections into a state machine requiring two independent signals and sustained evidence before alerting.
• Measurement:
• Hand-label ground truth from footage, including night shifts and full PPE.
• Measure false positive and negative rates per camera.
• Tune and document thresholds.
• Test explicitly across PPE, lighting, distance, and skin tone, reporting any disparity.
• Reporting:
• Develop an event store.
• Create a Grafana dashboard.
• Prepare a written evaluation with a clear recommendation (proceed, change approach, or stop).

Requirements & Tech Stack

• Strong Python.
• Computer vision fundamentals: detection, pose estimation, tracking.
• Proficiency with OpenCV and video concepts (codecs, frame rates, RTSP).
• Familiarity with PyTorch for running, fine-tuning, and evaluating models.
• Experience with Linux, ffmpeg, basic networking, and Git.
• Instinct to measure rather than assume.

Nice to Have

• TensorRT / ONNX Runtime
• GStreamer or DeepStream
• MediaPipe
• YOLO-family models
• RTMPose
• ByteTrack
• Grafana
• MQTT
• Exposure to industrial environments and SCADA/PLCs.

Qualifications

• Currently enrolled in or recently graduated from a program in Computer Science, Electrical/Electronic Engineering (EE/ECE), Robotics, or a similar field.
• Must be able to work on-site.

This is a paid internship.