Overview

About Biz-Tech Analytics

Responsibilities

• Design, develop, and deploy advanced Computer Vision models for real-world applications including object detection, object tracking, image classification, instance/semantic segmentation, pose estimation, human activity recognition, and visual quality inspection.
• Develop real-time video analytics pipelines capable of processing high-volume image and video streams from cameras and other visual sensors.
• Work with architectures and frameworks such as YOLO, Faster R-CNN, RetinaNet, ViT-based models, SAM, Grounding DINO, and other modern vision architectures.
• Build and optimize image and video processing pipelines, including frame extraction, preprocessing, augmentation, feature extraction, post-processing, tracking, and inference.
• Develop robust solutions for multi-object tracking, person detection, behavior/activity analysis, anomaly detection, and visual event recognition.
• Perform model optimization for production inference, including quantization, pruning, batching, model compression, GPU acceleration, and inference optimization where required.
• Build Computer Vision systems for both edge and cloud environments, balancing latency, accuracy, throughput, memory utilization, and scalability.
• Work with OpenCV, PyTorch, TensorFlow, CUDA, ONNX/ONNX Runtime, TensorRT, and other relevant Computer Vision technologies.
• Design and maintain scalable computer vision data pipelines, including image/video collection, dataset preparation, annotation workflows, data validation, and training-data quality improvement.
• Analyze model performance using appropriate Computer Vision evaluation metrics such as mAP, IoU, precision, recall, F1-score, tracking metrics, and inference latency.
• Debug and improve models that perform poorly in real-world conditions such as low-light environments, camera variations, occlusion, motion blur, changing backgrounds, and domain shifts.
• Lead the transition of Computer Vision PoCs into reliable production systems, ensuring models remain accurate and performant under real-world operating conditions.
• Mentor junior ML engineers on Computer Vision modelling, experimentation, debugging, optimization, and production deployment.
• Work directly with clients and internal stakeholders to understand visual AI problems, define technical approaches, develop PoCs, and translate them into scalable production solutions.
• Continuously evaluate emerging Computer Vision research, architectures, datasets, and foundation models and assess their applicability to production use cases.

Qualifications

• Strong expertise in Python and practical experience developing Computer Vision solutions using PyTorch, TensorFlow, OpenCV, or equivalent frameworks.
• Strong understanding of deep learning for Computer Vision, including CNNs, transformers, object detection, classification, segmentation, tracking, and representation learning.
• Proven experience building and deploying models for image and/or video-based applications.
• Strong hands-on experience with at least several of the following:
• Object Detection
• Multi-Object Tracking
• Image Classification
• Semantic / Instance Segmentation
• Pose Estimation
• Human Activity Recognition
• OCR / Document Vision
• Anomaly Detection
• Visual Quality Inspection
• Video Analytics
• Understanding of GPU-based inference and experience with technologies such as CUDA, TensorRT, ONNX Runtime, or similar optimization frameworks.
• Strong understanding of Computer Vision evaluation and error analysis, with the ability to diagnose model failures and improve performance.
• Good understanding of MLOps practices required to train, deploy, monitor, version, and maintain production Computer Vision models.
• Previous experience mentoring engineers or taking technical ownership of Computer Vision projects.
• Strong problem-solving skills and the ability to independently take a Computer Vision problem from concept → PoC → production.
• Exposure to Generative AI and Vision-Language Models (VLMs).
• Exposure to AI governance, data privacy, security, and compliance requirements for visual data.