Saif Ullah

Principal Data Scientist

Jan 2022 - Jul 2025 • 3 yrs 6 mos

Architected an enterprise RAG system processing 5000+ technical documents, MSDS sheets, and safety protocols, leveraging multi-modal architecture and custom embeddings on cloud-based platforms for 97% retrieval accuracy. Architected a multi-modal NLP system for automating the engineering review process of client contract clauses, utilizing historical data spanning 5+ years to train models and deliver precise technical feasibility assessments and compliance recommendations. Engineered an advanced information extraction pipeline for a European insurance provider that identified 50+ entity types using dynamic few-shot prompting with Claude 3.5 Sonnet on AWS SageMaker, achieving 96% accuracy and supporting scalable AI solutions. Developed a comprehensive BWA data extraction system for a financial institution using AWS Textract and NLP, creating a complex. carbon footprint calculation engine in accordance with European emission standards, which improved the accuracy of emission reporting. Developed an advanced AI-powered signal prediction system using deep learning to automatically classify and predict industrial signal attributes, enhancing operational efficiency by enabling intelligent categorization and predictive analytics for industrial operations. Led and managed a cross-functional team of 15+ professionals, including data engineers, data scientists, backend developers, and front-end engineers, which established robust MLOps practices and automated deployment pipelines, improving project delivery speed and enhancing stakeholder satisfaction.

Computer Vision Engineer | Python

Oct 2021 - Jul 2025 • 3 yrs 9 mos

Developed an advanced computer vision application using OpenCV and custom image processing algorithms to detect tiles in building images with 99.2% accuracy, incorporating edge detection, contour analysis, and morphological operations within a scalable framework. Implemented sophisticated tile segmentation and virtual repainting functionality, enabling real-time photorealistic modification of images and demonstrating effective prototyping and functional design skills. Engineered a custom algorithm for automatic cost estimation based on detected tile dimensions, selected paint types, and coverage requirements, delivering accurate real-time renovation cost calculations. Created optimized C++ scripts for core image processing and supported seamless integration with iOS platforms, ensuring high-performance execution in mobile environments and reliable API interfacing. Designed a robust image processing pipeline combining edge detection, contour mapping, dimensional analysis, and color transformation techniques to accurately identify tile boundaries, calculate surface areas, and apply realistic paint effects while maintaining original image quality.