AI/ML Engineer - Computer Vision / YOLO

AI/ML Engineer – Computer Vision
Location: Gurugram, Haryana
Work Mode: Remote / Hybrid / On-site
Experience: 1+ year relevant experience, or strong hands-on project/internship experience
Domain: AI/ML | Computer Vision | Construction Technology
About the Role
We are building AI-powered capabilities for construction quality and inspection and are looking for an AI/ML Engineer with practical experience in Computer Vision and YOLO-based model training.
You will work across the complete model lifecycle — from data collection and annotation to model training, evaluation, optimization, deployment, and production feedback.
This role is ideal for someone who enjoys building and improving models hands-on, rather than simply integrating third-party AI APIs.
Key Responsibilities
• Train, evaluate, and improve YOLO-based object detection models for construction defect detection.
• Prepare, clean, annotate, version, and manage image datasets collected from construction sites.
• Define annotation guidelines and defect classes with product and domain teams.
• Perform data augmentation, dataset balancing, train/validation/test splitting, and quality checks.
• Use frameworks such as Ultralytics YOLO and PyTorch for model development and experimentation.
• Evaluate models using Precision, Recall, F1 Score, mAP, confusion matrices, and class-wise performance.
• Reduce false positives and false negatives through dataset and model improvements.
• Explore object detection, classification, and segmentation depending on the use case.
• Optimize trained models for practical inference speed and production deployment.
• Build or support inference APIs/services for integration with mobile, web, and existing .NET systems.
• Maintain experiment notes, model versions, checkpoints, and reproducible training configurations.
• Research new computer vision approaches that can improve construction inspection automation.
Required Skills & Qualifications
1+ year of relevant AI/ML experience, or strong hands-on project/internship experience in computer vision.
Good Python programming skills.
Hands-on experience training YOLO models.
Understanding of object detection and image classification concepts.
Experience with PyTorch and/or TensorFlow.
Experience preparing and annotating image datasets.
Working knowledge of OpenCV.
Understanding of training, validation, overfitting, augmentation, confidence thresholds, and model evaluation.
Basic familiarity with GPU-based training and model deployment.
Basic understanding of REST APIs for serving model inference
Good to Have
• Experience with YOLOv8, YOLO11, or similar architectures.
• Knowledge of image segmentation and instance segmentation.
• Experience in construction, manufacturing, industrial inspection, or visual quality-control use cases.
• Experience with CVAT, Roboflow, Label Studio, LabelImg, or similar annotation tools.
• Knowledge of ONNX and model optimisation for mobile or edge deployment.
• Experience with NVIDIA GPUs, CUDA, or cloud GPU platforms.
• Knowledge of FastAPI or Flask.
• Understanding of MLOps, experiment tracking, and model versioning.
• Awareness of vision-language or multimodal AI models.
Example AI Use Cases
You may work on Computer Vision models for identifying:
Concrete cracks and surface damage
Honeycombing and poor concrete finishing
Tile, painting, and plastering defects
Leakage or dampness indicators
Visible reinforcement and workmanship issues
PPE and safety compliance
Installation and material defects
Other project-specific quality issues identified from construction site images
What We Expect
We are specifically looking for candidates who understand the complete model-building lifecycle:
Data Collection → Annotation → Dataset Preparation → Training → Evaluation → Improvement → Deployment → Production Feedback
Candidates who can confidently explain and practically apply concepts such as dataset quality, overfitting, augmentation, mAP, confidence thresholds, class imbalance, false-positive/false-negative reduction, and model improvement strategies will be preferred.
Who Will Be a Good Fit?
You will be a strong fit if you:
Like working hands-on with Computer Vision models and datasets.
Have actually trained and evaluated YOLO models.
Understand why a model performs poorly and how to improve it.
Are comfortable working with imperfect, real-world image data.
Can move beyond experimentation and think about production deployment and inference performance.
Enjoy solving domain-specific problems where the data and use case are not always straightforward.
Work Mode
Flexible — Remote / Hybrid / On-site
Location: Gurugram, Haryana

Work arrangement
No