Khurram Shafiq

Senior Data Scientist

Government of Canada • Temps plein

Mar 2024 - Present • 2 yrs 2 mos

Led the development and deployment of LLM-powered solutions (LLaMA, Mistral, Claude, BGE) on Databricks, building RAG-based AI agents using LangChain for real-time diagnostics and reliability insights. Designed and implemented time-series anomaly detection models (LSTM, Transformers) for critical assets including PHT pumps, bearings, turbines, and hydrogen leak modeling, enabling predictive maintenance. Developed computer vision solutions, including fuel bundle detection modules and industrial monitoring systems (e.g., TRF gauge readings). Built scalable data pipelines and MLOps systems using Azure Data Factory, PySpark, Delta Lake, and MLflow for production-grade deployments. Created OCR-driven document intelligence systems (LayoutLMv3, Tesseract) and advanced analytics dashboards (Power BI) for operational insights. Contributed to digital twin models and forecasting solutions, including trash rack performance forecasting, enhancing asset reliability and decision-making. Applied zero-shot learning, embeddings, and AI agent frameworks to improve model generalization and automation across complex industrial systems.

Machine Learning Engineer

Fintech • Temps plein

Apr 2022 - Feb 2024 • 1 yr 10 mos

Designed and deployed LLM-powered systems (BERT, LayoutLM, LXMERT, GPT-based models) and AI agents using LangChain and LlamaIndex for document validation, transaction intelligence, and compliance automation. Built scalable RAG pipelines, embeddings, and vector search systems for extracting and processing financial and transactional data. Developed OCR and document intelligence solutions using LayoutLMv3, Google Document AI, and multimodal models for invoice and document verification. Engineered real-time ML systems for payment routing and transaction classification using Kafka, streaming pipelines, and transformer-based architectures. Implemented computer vision models for KYC, including facial recognition and image retrieval systems using vector databases. Built end-to-end ETL pipelines and big data architectures (Databricks, Spark, Hadoop) to process multi-terabyte datasets. Developed predictive models for transaction timing, classification, and revenue forecasting, enabling data-driven decision-making. Applied zero-shot learning, embeddings, and semantic modeling to improve adaptability across new transaction types. Established MLOps and observability frameworks using MLflow, Docker, Kubernetes, CI/CD, and monitoring tools for scalable production deployment. Integrated model explainability (LIME, SHAP) and human-in-the-loop systems to enhance trust, accuracy, and continuous learning.

Machine Learning Engineer (Computer Vision)

UtilityNFTCoin • Temps plein

Nov 2020 - Mar 2022 • 1 yr 4 mos

Developed advanced computer vision algorithms using OpenCV, Halcon, Cognex, and Python for object detection, comparison, and replication tools in engineering systems. Applied image processing, segmentation, and denoising autoencoders to enhance and clean technical drawings. Implemented feature extraction (SIFT, ORB, BRISK) and homography techniques for image alignment and de-skewing. Built OCR pipelines using Tesseract, EasyOCR, and custom deep learning models to extract text and identifiers from engineering drawings. Designed context-aware AI systems combining computer vision and NLP (model fusion) to extract drawing numbers and equipment codes. Developed deep learning models using TensorFlow, PyTorch, and Keras for image analysis and optimization tasks. Created hybrid recommendation systems (collaborative filtering + deep learning + NLP with RoBERTa) for intelligent content matching. Built and deployed ML applications and APIs using Flask/Django, with scalable infrastructure on Azure and GCP. Optimized performance using GPU acceleration, Cython, and Numba for real-time processing. Implemented MLOps and deployment workflows using Docker, Kubernetes, CI/CD, and version control (Git). Contributed to R&D initiatives, delivering a comprehensive engineering drawing toolkit and supporting large-scale ML deployment strategies.