Khurram Shafiq
Toronto,Canada
Summary
With over 8 years of specialized experience in machine learning, AI, and computer vision, I have mastered a range of technologies including Python, TensorFlow, PyTorch, and OpenCV, underpinned by a solid foundation in data science principles and practices. My proficiency extends to leveraging large language models (LLMs) and developing sophisticated deep learning algorithms for innovative solutions in computer vision and predictive modeling. I've implemented cutting-edge frameworks such as Keras, Django, and Flask, and engaged in advanced NLP techniques to enhance AI applications. My work encompasses the entire ML lifecycle, from data preprocessing and model development to deployment and MLOps, utilizing cloud platforms like AWS and Google Cloud for scalable, efficient solutions. My expertise in CI/CD pipelines, combined with a keen insight into pattern recognition and segmentation analysis, positions me at the forefront of AI technology innovation.
Overview
8
8
years of professional experience
1
1
Certification
Work History
Machine Learning Engineer
Buckzy Payment Inc
Toronto, Canada
04.2022 - Current
- Developed a documentation and invoices validation tool utilizing LayoutLM, LayoutLMv3, and Google Document AI, enhancing accuracy and efficiency in processing critical documents.
- Collaborate with stakeholders to understand business requirements and design and deploy scalable production LLM pipelines, ensuring seamless integration with existing systems.
- Designed and implemented a comprehensive solution for generating client statements using ETL (Extract, Transform, Load) processes and Databricks platform.
- Implement entity recognition models using sensor fusion, contributing to the accurate extraction of critical information from transaction remarks.
- Implemented a facial recognition model utilizing state-of-the-art AI techniques in the vision space for KYC (Know Your Customer) processes, and further enhanced the system by optimizing an image retrieval model using vector databases to ensure efficient and accurate matching of customer images.
- Developed and implemented end-to-end data pipelines using Python and ETL tools to handle large, multi-terabyte datasets.
- Implement various cloud-based ingestion tools for compliance and developer portals, streamlining data management processes.
- Develop and implement document verification using OCR technologies to enhance compliance processes.
- Implement whitelisting and blacklisting of transaction remarks using advanced machine learning models, leveraging LLM models, langchain, and llamaIndex for fine-tuning, transfer learning, and accuracy improvements.
- Developed and deployed a machine learning model utilizing to accurately classify transaction statuses and predict transaction times. The model leveraged contextual embeddings and transformer-based architectures to capture semantic meaning and temporal dependencies within transaction data.
- Implemented real-time rule engines and optimized transactions, removing dependencies for compliance in real-time scenarios.
- Implemented zero-shot learning in transaction models, enabling accurate predictions for new transaction types without explicit training. Leveraged semantic embeddings and auxiliary information to enhance adaptability and efficiency.
- Leveraged LXMERT, a powerful multi-modal framework, for OCR tasks combining text and image data. Integrated Google's "BERT" and "ResNet" models to accurately extract and recognize text from images, enabling efficient document validation processing.
- Integrated a predictive ML model for payment routing with Apache Kafka, enabling real-time transaction processing. Optimized transaction efficiency and reduced costs by applying the model on streaming data for dynamic routing decisions.
- Created an AI-powered revenue forecasting model that utilized advanced machine learning techniques to generate accurate revenue forecasts.
- Implement and deploy core algorithms in Python, utilizing natural language processing frameworks such as PyTorch, TensorFlow, AllenNLP, BERT, FastText, GloVe, Elmo, and the HuggingFace transformers library.
- Integrated DevOps practices into machine learning workflows, ensuring robust CI/CD pipelines for model deployment and updates. Utilized Docker for containerization and Kubernetes for orchestration, achieving scalable and efficient model deployment in cloud environments such as GCP.
- Implemented automated testing and monitoring to maintain model accuracy and performance.
- Continuously monitored and improved deployed models through post-deployment monitoring and continual learning capabilities, ensuring optimal performance and accuracy.
- Identified opportunities to enhance user experiences by leveraging LLM features and machine learning techniques, driving data-driven solutions that cater to the needs of customers and the organization.
- Deployed and maintained backend infrastructure in cloud environments (e.g., GCP) using Docker containers, managing NLP-based deep learning applications.
- Developed a revenue forecasting model using ETL processes and Tableau. Extracted and transformed data from various sources, then loaded it into Tableau for analysis.
- Created predictive models and interactive dashboards in Tableau, enhancing revenue prediction accuracy and facilitating data-driven decision-making.
- Collaborated closely with team members to develop machine learning systems from prototyping to production, contributing to the overall success of projects.
- Implemented a Machine Learning Observability Model by fine-tuning Integrated AI Cloud and leveraging the Llama Index for dynamic model adjustments. This system is specifically tailored for monitoring and enhancing the performance of payment transaction systems.
- Integrated tools like LIME and SHAP for in-depth interpretability and explainability, enabling stakeholders to understand model predictions and influence on transaction outcomes, thus fostering trust and transparency in automated financial decisions.
- Utilized Human-in-the-Loop Systems to incorporate expert feedback into the continuous learning cycle, ensuring that the model adapts to new patterns and anomalies in financial transactions while maintaining high accuracy and compliance standards.
- Developed and deployed a robust, scalable ML observability infrastructure, capable of handling large volumes of payment transactions, while providing real-time performance metrics, anomaly detection, and automated alerts for system deviations or potential fraudulent activities.
- Deployed and meticulously implemented state-of-the-art ML algorithms using MLflow, Databricks, DagsHub, and EvidentAI to establish a robust MLOps framework. This comprehensive approach enables version monitoring, real-time performance tracking, efficient data processing and analytics, algorithm implementation, and fosters enhanced collaboration while extracting valuable data insights.
- Implemented a robust data architecture for efficient big data management using Spark, Hadoop, distributed caching, MLFlow, and in-memory frameworks. This architecture enables real-time recommendations for payment routing using new features, enhancing user experiences and data-driven decision-making.
Machine Learning/Computer Vision Programmer
Kinectrics
Toronto, Canada
11.2020 - Current
- Implemented core algorithms for image processing with OpenCV-Python, OpenCL, or other 3rd party libraries (Halcon, MIL, Cognex) for Search Object Tool, Comparison Tool, Replication Tool etc.
- Working with computer vision morphological & segmentation techniques and denoising auto-encoders in python to remove noise in technical drawings.
- Worked on feature extraction (using the orb, sift and brisk algorithms in python) and homography for de-skewing and alignment for engineering drawings.
- Optimizing technical drawings with machine learning algorithms for enhancing system applications.
- Working in Python, TensorFlow, PyTorch, and Keras for deep learning applications.
- Implemented a computer vision OCR algorithm in OpenCV-Python, Tesseract, and EasyOCR (Python) to extract the letters and digits. Also, training a custom model to enhance the accuracy using annotated images and text.
- Implemented an advanced recommendation engine using a hybrid approach, combining deep learning and collaborative filtering. Employed a neural collaborative filtering model that integrates a multi-layer perceptron for complex pattern recognition with user-item interaction data.
- Enhanced the model's contextual understanding through NLP techniques, utilizing transformer models like RoBERTa for content analysis. This resulted in highly personalized and relevant recommendations, adapting dynamically to user preferences through feedback loops.
- Implemented a computer vision smart contextual learning algorithm in OpenCV-Python with NLP text analysis (Model Fusion) to extract drawing numbers & equipment codes etc.
- Responsible for documentation of software applications, verification plans, release memos and user manuals.
- Developed a content-based recommendation system utilizing TF-IDF (Term Frequency-Inverse Document Frequency) and cosine similarity algorithms to analyze and match user profiles with relevant content.
- Implemented a hybrid model combining collaborative filtering with matrix factorization (using Singular Value Decomposition - SVD) to capture both user-item interactions and content features, significantly improving recommendation relevance and personalization.
- Implemented a graphical user interface (GUI) for computer vision applications using Tkinter-Python etc.
- Implementing & deploying the trained models on web services using Django and Flask in the Python framework.
- Implemented projects and managed version control in GitHub and Bitbucket using Git.
- Working in cython, numba and GPU parallel computing to optimize code in real time in Python.
- Contributed and worked with the company’s R&D department to implement innovative ideas for engineering drawing toolkit for software applications.
- Working with containers like Docker & Kubernetes for the deployment phase.
- Worked briefly in a team which was responsible for a roadmap of ML deployment using Google Cloud infrastructure for software development.
- Implemented a Flask API for image processing with diverse ML models, adapting to evolving needs. Deployed a scalable infrastructure with real-time monitoring on Azure using IIS for enhanced efficiency and reliability.
Machine Learning Engineer (Computer Vision)
Extend AI
Ottawa, Canada
01.2021 - 03.2022
langchainLayoutLMAutoTSDynaconfSocket.ioDataStageDatabricksPowerBIMLOpsMLOpsKerasKerasKeras
- Developed AI and machine learning systems in Python for predictive modeling, including anomaly and object detection, enhancing algorithm accuracy for user recommendations.
- Implemented anomaly detection, object detection, and image classification algorithms using Python, PyTorch, TKerasFlow, and Keras.
- Utilized GANs (FineGAN, CycleGAN, StyleGAN) for texture and image analysis with OpenCV and Python.
- Developed image processing and feature extraction algorithms with OpenCV and other libraries for stitching and color correction.
- Generated synthetic data with Computer Vision and Deep Learning techniques, including Defect GAN and StyleGAN2.
- Applied ONNX compression for ML models to leverage C++ computational efficiency in Python.
- Managed data pipelines for ML model training, testing, and validation, assessing new data sources for client projects.
- Deployed custom deep learning models (CNN, RNN, LSTM, etc.) on large datasets, using advanced frameworks like Facebook AI's DINO.
- Implemented ML model deployment on edge devices, such as AGX and Nvidia Jetson, and utilized Raspberry Pi camera modules for auto-calibration with OpenCV and machine learning.
- Worked in a fast-paced, agile software development environment, conducting literature reviews on state-of-the-art model deployment.
Software Developer
Mitacs
Ottawa, Canada
03.2021 - 08.2021
- Developed a chatbot at Spreevel Inc., sponsored by Mitacs, utilizing NLP models such as BERT, GPT-2, GPT-3 (OpenAI), and Hugging Face.
- Implemented BERT for chatbot functionality and worked on back-end services using an entity framework in C.
Software Developer (Machine Learning & AI Researcher)
National Research Council
Ottawa, Canada
01.2019 - 12.2020
- Collaborated with QinetiQ and CS Canada on a federal prison project, focusing on UAV object detection using deep neural networks like CNN and R-CNN.
- Optimized UAV and bird detection models using YOLO and R-CNN, enhancing accuracy and performance.
- Developed custom AI frameworks for classification, object detection, and segmentation to meet project specifications.
- Utilized UDP/IP and TCP/IP for socket programming in Python, enabling efficient real-time application communication.
- Designed and tested a real-time software integrating radar and optical sensors with fusion algorithms for enhanced detection capabilities.
- Trained and deployed advanced deep learning models (CNN, RNN, LSTM) on extensive datasets for precise object identification.
- Implemented LSTM models and employed Kalman filtering and deep association matrices (e.g., deep sort) for robust object tracking, complemented by IMM Radar tracking for feature extraction.
- Applied reinforcement learning techniques using MAB (Multi-Armed Bandit) and MDP (Markov Decision Process) models to refine learning algorithms.
- Engineered and documented a distributed Python system, leveraging multiprocessing and multithreading for real-time object tracing.
- Achieved a 70% performance improvement in real-time software operation through strategic optimizations.
Quality Manager
Kristofoam Industries Incorporation
Toronto, Canada
01.2018 - 01.2019
- Worked closely with upper management to define quality benchmarks, aligning product standards with company objectives.
- Conducted comprehensive testing to verify final product compliance with established quality benchmarks, reinforcing adherence through regular audits.
- Developed and implemented robust quality control processes, guaranteeing consistent fulfillment of quality criteria.
- Documented inspection results and communicated key insights to the production team for continuous improvement.
- Led a dynamic production team across two shifts, overseeing approximately 40 employees to ensure efficient operations and high-quality output.
Junior Engineer
Techno Consult International
Lahore, Pakistan
01.2017 - 09.2017
- Assisted the Advisor Transmission and consultants on US $40 Million Renewable Energy Projects by MWH Americas Inc., backed by USAID, enhancing monitoring, inspection, and design review processes.
- Developed comprehensive electrical designs and drawings for Residential Solar PV systems using AutoCAD and specialized design tools, ensuring readiness for permitting and installation.
- Optimized project designs by analyzing constraints, improving overall performance and efficiency.
- Conducted commissioning and performance analysis of Solar Power systems to ensure optimal operation.
- Performed site visits for initial surveys and to implement Quality Assurance protocols, ensuring project standards were met.
- Investigated failure mechanisms and implemented strategies for yield improvement, enhancing system reliability and efficiency.
Project Engineer
Avotech Engineering Services
Lahore, Pakistan
08.2016 - 01.2017
- Developed expertise in 3-phase AC/DC rectifiers and DC/AC inverters, ensuring efficient energy conversion processes.
- Specialized in Variable Speed Drive systems and Active Power Filter systems, enhancing operational efficiency and power quality.
- Designed L-C-L filters for grid-connected converters, adhering to CSA, UL, and CE design standards and requirements.
- Gained hands-on experience with IGBT and SiC MOSFET devices, leveraging advanced semiconductor technologies for improved system performance.
Education
Master of Computer Science & Concentrated Applied Artificial Intelligence - Machine Learning & Artificial Intelligence
University of Ottawa
Bachelor of Science in Electrical & Computer Engineering -
University of Engineering & Technology
Exchange Program for One semester in Bachelor of Science - Cultural Exchange Program in United States of America
Saginaw Valley State University
Skills
- Skilled in Python, Keras, PyTorch, TensorFlow, OpenCV; proficient in Java, C, C#; experienced with Spark, MongoDB, Hadoop
- Versatile in data mining, QA, BI with Tableau, PowerBI; adept in database management, SQL; knowledgeable in SAS, Databricks, ETL with IBM DataStage
- Expert in deep learning, computer vision, machine learning; skilled in NLP, LLMs, MLOps; experienced with CI/CD, Jenkins, ML Flow
- Proficient in cloud platforms (AWS, Azure, Google Cloud); skilled in Docker, Kubernetes; experienced with Linux, Windows, Eclipse, Visual Studio
- Adept in cloud-native applications, API Gateways, Socketio; knowledgeable in Git, GitHub, data versioning with Delta Lake, configuration with Dynaconf
- Experienced in pattern/trend identification, high-volume data systems; skilled in Math (statistics, calculus, linear algebra, probability)
- Proficient in sensor fusion, zero-shot learning, segmentation; experienced with forecasting, prediction models, AutoTS; adept in training and prediction pipelines in MWAA, predictions in Snowflake
Accomplishments
- Cultural Ambassador of Pakistan in USA: This is the most appreciable academic scholarship given to exceptional students at the undergraduate level by GLOBAL UGRAD Pakistan to recognize them on global platforms.
- The Golden Key Award: This medal is awarded to the most outstanding graduating student in Electrical Engineering•
- Dean’s List –University of Engineering & Technology, PAK: The purpose of this list is to recognize exceptional students of the engineering and computer science faculties who have distinguished themselves by their outstanding academic achievements. It honors the top 15% of the students.
Conferences Publications
AI Enabled Drone vs Bird Detection, Classification & Sensor Fusion on AUVSI XPONENTIAL 2020 (Association for Unmanned Vehicle Systems International), https://ruor.uottawa.ca/handle/10393/45474
Certification
- Bruce Power security clearance (Level-3)
- OPG Security clearance
- Cultural Amassador of Pakistan in USA
Languages
English
Native/ Bilingual
References
References available upon request.
Timeline
Machine Learning Engineer
Buckzy Payment Inc
04.2022 - Current
Software Developer
Mitacs
03.2021 - 08.2021
Machine Learning Engineer (Computer Vision)
Extend AI
01.2021 - 03.2022
Machine Learning/Computer Vision Programmer
Kinectrics
11.2020 - Current
Software Developer (Machine Learning & AI Researcher)
National Research Council
01.2019 - 12.2020
Quality Manager
Kristofoam Industries Incorporation
01.2018 - 01.2019
Junior Engineer
Techno Consult International
01.2017 - 09.2017
Project Engineer
Avotech Engineering Services
08.2016 - 01.2017
Master of Computer Science & Concentrated Applied Artificial Intelligence - Machine Learning & Artificial Intelligence
University of Ottawa
Bachelor of Science in Electrical & Computer Engineering -
University of Engineering & Technology
Exchange Program for One semester in Bachelor of Science - Cultural Exchange Program in United States of America
Saginaw Valley State University
Similar Profiles
ANISH. PANISH. P
Manager - Operations & Customer Support at Buckzy Payments LtdManager - Operations & Customer Support at Buckzy Payments Ltd
Ayan KhatibAyan Khatib
Technical Account Manager at Razorpay-Payment Gateway & Payment SolutionsTechnical Account Manager at Razorpay-Payment Gateway & Payment Solutions
Crystal HadenCrystal Haden
Merchant Services Sr Acct Rep at Payment Alliance/Clearent Payment ProcessingMerchant Services Sr Acct Rep at Payment Alliance/Clearent Payment Processing
Jenny Cheng LYUJenny Cheng LYU
Salad Prep Cook (Cold Station) at Cafe on the SquareSalad Prep Cook (Cold Station) at Cafe on the Square
AJIKE SOMUYIWAAJIKE SOMUYIWA
Business Analyst/ Project Associate at The Marketing StoreBusiness Analyst/ Project Associate at The Marketing Store