AI Intern Computer Vision

Job Title:

Location:

Start Date:

Role Overview

As an AI Intern, youll work closely with senior AI engineers on developing and deploying state-of-the-art deep learning models for video-based human activity analysis, pose estimation, and motion understanding. This is a hands-on role focused on building and integrating real-world AI systems, not academic research. Youll get the chance to work on projects involving posture correction, rep counting, and exercise classification, using tools and frameworks like YOLOv8, Vision Transformers, MMAction2, and Mediapipe.

Key Responsibilities:

• Design, train, test, and fine-tune deep learning models for human motion understanding, classification, and detection using video data.

• Apply transfer learning and fine-tuning on state-of-the-art (SOTA) models like YOLO, ResNet, EfficientNet, Vision Transformers, and UNet.

• Build and optimize real-time computer vision pipelines in Python using PyTorch and/or TensorFlow.

• Implement pose estimation, skeleton tracking, and posture correction models using libraries such as Mediapipe or OpenPose.

• Apply camera calibration techniques and geometric corrections for accurate visual measurements.

• Preprocess large-scale video datasets, augment data for robustness, and manage annotation workflows.

• Integrate AI modules with REST APIs or cloud systems (optional), and assist in model packaging using ONNX, TensorRT, or TorchScript.

• Collaborate with mentors and team members to align design, timelines, and delivery.

Required Skills:

Proficiency in Python, with clean, modular code and understanding of performance optimization.

• Experience with PyTorch and/or TensorFlow.

• Hands-on experience working with:

• Object detection (YOLOv5, YOLOv8)

• Video-based models (SlowFast, I3D, TimeSformer)

• CNN architectures (ResNet, EfficientNet, Vision Transformers)

• Pose estimation frameworks (Mediapipe, OpenPose)

• Strong grasp of camera calibration, 3D geometry, and motion analysis. • Familiarity with OpenCV, NumPy, and image/video preprocessing

. • Comfortable integrating and adapting open-source models and APIs.

Preferred Skills:

• Exposure to tools like MMAction2, Detectron2, or DeepSort.

• Understanding of human activity recognition, especially in fitness or healthcare contexts.

• Experience deploying models using ONNX, TensorRT, or similar inference engines.

• Familiarity with cloud services (Azure, AWS) and model integration through REST APIs.

• Knowledge of real-time edge inference or embedded systems is a plus. What Youll Gain

• Mentorship from seasoned AI engineers in a high-impact production environment.

• Experience working on end-to-end AI pipelines in a fast-growing sector.

• Exposure to real datasets, users, and performance-critical deployment.

• Opportunity to receive a pre-placement offer (PPO) based on internship performance.