I am a dedicated and analytical Machine Learning Engineer with extensive experience in developing and deploying NLP and classification models. My background includes hands-on work with sentiment analysis, text classification, and large language models (LLMs), such as Meta LLAMA-2 and Phi-3.5, all of which are crucial for understanding and analyzing data.
I possess a strong foundation in data engineering and machine learning, demonstrated through various projects involving text summarization, classification, and conversational AI. My expertise in leveraging transformer-based models allows me to extract meaningful insights from complex data sets.
I am committed to continuous learning and staying updated with the latest research in machine learning. I thrive in collaborative environments and am eager to contribute to projects that require technical proficiency and innovation.
CGPA: 3.90
Core Courses: Advanced Algorithms Analysis, Mathematical Methods for Computing, Theory of Computation, Operating Systems
Elective Courses: Machine Learning, Deep Learning, Data Mining, Natural Language Processing (NLP)
Marks: 80.58%
Courses: Data Structures, Algorithms, Databases, Networking, C++, E-Commerce, Digital Logic Design, Software Engineering, Compiler, Artificial Intelligence, Operating Systems
Courses: Computer Science, Mathematics-A, Physics
Developed a voice-based AI assistant system designed to support elderly individuals by answering queries related to their health, medicines, and other relevant topics. This system utilized the Phi-3.5 model and integrated various open-source text-to-speech and speech-to-text models to create a comprehensive pipeline. The assistant functions similarly to Siri, providing voice-activated responses to help users manage their daily needs and inquiries.
Developed an advanced document summarization system using Meta LLAMA-2 integrated into a Django-based web app. This project, conducted in a secure environment with limited resources, involved summarizing documents of varying lengths and formats—from a few sentences to hundreds of pages. Leveraging a quantized Meta LLAMA-2 13B model and collaborating closely with Django and DevOps teams, I ensured efficient processing and accurate summaries tailored to specific formats and requirements.
Developed a domain classification system for a complex document dataset with nearly 100 domains. Overcame challenges of limited and noisy labeled data by generating synthetic data and fine-tuning roBERTa base models. Designed a hierarchical system with 16 parent models for initial broad domain classification and various child models for subdomain classification, achieving around 80% accuracy.
Developed a Sentiment Analysis model to classify paragraphs within documents into five categories: negative, extreme negative, neutral, positive, and extreme positive. This model provided detailed sentiment insights for document content, enhancing the ability to understand and interpret textual data effectively.
Designed and implemented a data processing pipeline system for the Azadea Group, including Azadea, Decathlon, Adidas, etc. The system handled over 10GB of XML files daily, generating multiple CSV and XML files. Faced with the challenge of limited hardware (8GB RAM on a Windows laptop), I utilized creative solutions to sequentially process data, avoiding memory overload.
Developed "Message Health Checker," a sophisticated roBERTa-based classification system for LinkedIn messages. This tool categorizes messages into 5 multi-label categories with 5 ranks each, assessing tone and effectiveness. It empowers users to craft highly personalized messages, boosting engagement by 10x and tracking campaign success with detailed conversion metrics.
Developed a multiclass multilabel text classification model for job advertisement messages sent by recruiters on LinkedIn to job candidates. The classification determines the quality of the message and uses an AI-based text assistant. This project is a component of QLU.ai's primary product.
Created a hybrid approach algorithm to split job descriptions into logical sections. This component works as an assistant to improve the precision and recall of other components of QLU.ai's primary product.
Trained transformer-based models such as BERT, RoBERTa, and XLNet to classify each section of a job description. This helper component increases the accuracy of other components of QLU.ai's primary product, thereby enhancing customer trust in the product.
Researched pre-trained models with good accuracy for paraphrasing, text summarization, and grammar correction. These models serve as helper components for QLU.ai's message generation feature.
Developed a conversational AI system to handle communication between recruiters and candidates on LinkedIn.
Academic Project: Applied Feed-Forward Neural Network (FFNN) and Convolutional Neural Network (CNN) to classify Pokémon images into different categories.
Academic Project: Developed a text generation model based on RNN to generate names.
Academic Project: Trained Skip-gram, CBoW, and SVD models on a custom dataset to generate meaningful word embeddings.
Deployed various services on cloud hosting platforms like AWS, GCP, and Azure. Developed scalable data pipelines using AWS and Azure services. Utilized EC2 instances with external FTP, S3 connected to AWS Lambda for data processing, and Azure Virtual Machines with Azure Functions linked to Blob Storage for efficient data handling. Integrated GitHub Actions as a CI/CD pipeline to automate deployment and streamline workflows across both cloud environments.
Secured 15th position among the Top 20 students from Government Colleges in BISE Peshawar.
January 2014
GAT is taken by all graduated students in Pakistan. The GAT Test is a requirement for applying for a Master's Degree (MS) in Pakistan. Scored 83 out of 100 marks with a 99.62 percentile.
January 2019
Natural Language Processing Specialization on Coursera [L9XZX72XDF2A]
Natural Language Processing with Classification and Vector Spaces [HV2BR2GHE7SU]
Natural Language Processing with Probabilistic Models [9ZKN4QEUBSDC]
Natural Language Processing with Sequence Models [Z84QTPRB9C6L]
Natural Language Processing with Attention Models [7SYPZ5D8DPDN]
Introduction to Large Language Models [LEKT5FXJHJ56]
Generative AI with Large Language Models [RBXM49D5MZH4]
Build Basic Generative Adversarial Networks (GANs) [Z7LPD7T6M3A4]
Introduction to Self-Driving Cars [5AYN5M4FG8LW]
What is Data Science? [XDT4LQRWLGNB]
Python Essentials for MLOps [96WYAFV6J2WJ]
AI For Everyone [46G53XPMEA6S]
How Google does Machine Learning [G3HM7DJ4XG5Z]
Artificial Intelligence on Microsoft Azure [2Q55W93P5JPL]
Introduction to Deep Learning [UF89T3S2QVQG]
Neural Networks and Deep Learning [LVB52JUQ9HU7]
Supervised Machine Learning: Regression and Classification [Y7ECGZDYCVWY]
Mathematics for Machine Learning: Linear Algebra [RYF7JRYNYMJS]