Computer Vision Engineer

Karachi, Pakistan

• • Engineered an end-to-edge automated weight estimation system by integrating and fine-tuning open-source Computer Vision models reducing manual weighing labour by 100% on deployed farms.
• • Deployed, tested, and optimized the weight estimation pipeline across 7+ farms and 30+ individual poultry sheds, tailoring model coefficients for different camera angles, chicken genders, and breeds (breeder & broiler).
• • Developed an automated data logging system that captures images, performs analysis, and records key metrics to structured Excel reports every 10 minutes, generating over 140 daily data points per shed for operational analysis.
• • Optimized segmentation and depth estimation models for CPU inference, reducing processing time by approximately 50% and enabling stable deployment on low-resource farm hardware.
• • Spearheaded the end-to-end development and deployment of a Biosecurity Human-Vehicle access control system by collaborating with R&D on model development, Hardware team on camera integration, Business stakeholders on requirements, Deployment teams on rollout, and Analytics on data utilization, ensuring seamless alignment from concept to production.
• • Developed a production-ready facial recognition system from research to deployment, comprising a database encoding pipeline for a farm’s personnel, real-time FaceNet-based recognition, and physical gate control integration, reducing reliance on manual security checks.
• • Built robust modules for real-time facial recognition processing on both standard webcams and RTSP/ONVIF video streams from IP cameras, achieving sub-second identification times.
• • Converted and quantized PyTorch models to ONNX, OpenVINO, FP16, and INT8 format for deployment on dedicated edge AI hardware, overcoming memory and processing constraints to run complex models on-device.

Karachi, Pakistan

• • Pioneered an AI-driven defect detection system for concrete infrastructure using Conditional GANs, incorporating advanced techniques for crack detection, quantification, and categorization.
• • Engineered high-performance optimizations in GAN models, achieving real-time inference (0.3 seconds average), significantly enhancing on-site inspection efficiency.
• • Designed and developed custom evaluation metrics for computer vision applications, encompassing both pixel-level and block-wise segmentation accuracy for comprehensive model assessment. s
• • Created a proprietary algorithm for real-time, pixel-level measurement of crack dimensions, advancing structural integrity assessments in infrastructure inspections.
• • Constructed an advanced Mask2Image generative AI model, enhancing segmentation capabilities for concrete crack images.
• • Engineered a novel hybrid GAN framework for predicting crack propagation under load in concrete structures, pushing the boundaries of predictive maintenance in civil engineering.
• • Engineered an advanced crack propagation prediction system using DCGAN-based techniques and latent space manipulation, enhancing structural damage forecasting capabilities.

Karachi, Pakistan

• • Conducted comprehensive research on Structural Health Monitoring (SHM) systems and Bridge Deterioration Models, contributing to the advancement of predictive maintenance strategies in civil infrastructure. .
• • Developed immersive Virtual Reality (VR) visualizations for virtual bridge inspections and damage propagation simulations, enhancing training procedures and risk assessment capabilities in infrastructure management.

Remote

• • Implemented and optimized Traditional ML algorithms and techniques for data analysis.
• • Collaborated in the development of predictive models for road traffic collision severity.