Uzair Adamjee

- Designed and executed large-scale evaluation workflows for LLMs, analyzing model responses across coding, analytical, and research-focused tasks. Improved reasoning reliability through structured feedback loops, bug detection, and minimal-change code fixes. Built Python-based data pipelines for debugging, EDA, and statistical validation using Pandas, NumPy, and Jupyter. Collaborated with research teams to enhance model performance, training data quality, and benchmark reliability. - Led a team of annotators to evaluate LLM-generated Python code for MLE-Bench competitions. Identified execution-blocking bugs, documented root causes, implemented minimal code fixes, and validated outputs. Ensured annotation consistency, high-quality reviews, and reproducibility across complex code submissions. - Tested LLMs with full Python interpreter capabilities (REPL) to perform end-to-end analytics tasks including EDA, data cleaning, regressions, visualizations, and business insights generation. Evaluated correctness of statistical reasoning, model usage, data transformation steps, and visualization outputs.

- Built AI content optimization tools for automated captions, highlight extraction, and player summaries (NLP + Vision models). - Developed AI matchmaking logic recommending opponents and partners using performance analytics, historical data, and ranking behavior. - Created AI-powered player insights including strengths/weaknesses detection, performance patterns, and rating progression models.

- Create a Streamlit application that leverages Google Generative AI API to assess and score resumes for Applicant Tracking Systems (ATS). The tool automates the evaluation of resumes by analyzing key elements like keywords, skills, and formatting, providing an overall ATS-friendly score. - Developed a chatbot using OpenAI GPT-3 and LangChain to provide instant customer support by querying a database of FAQs and manuals, significantly reducing response time. - Developed a system to analyze and summarize legal documents, enabling lawyers to quickly grasp key points and saving significant review time using OpenAI. - Developed an app that leverages generative AI models to create graphs and charts based on user input. The app integrates with multiple LLMs, including Hugging Face, Ollama, and GPT, allowing users the flexibility to choose their preferred model for generating visual outputs. This solution provides a seamless interface where users can input data or instructions, and the AI models generate dynamic, visually appealing charts and graphs based on the data.

* Developed a machine learning classification model utilizing a dataset of approximately 20,000 articles extracted from a client-specific database. The project involved cleaning and preprocessing the data to ensure it was suitable for machine learning algorithms. Various machine learning (ML) and deep learning (DL) algorithms were applied to evaluate the model's performance, with a focus on F1-Score metrics. Technologies employed in this project included Python and SQL. * In the quest for innovation and improvement, meticulous manual tagging of Arabic data is being undertaken to construct a bespoke dataset. Leading a highly skilled team, an advanced Named Entity Recognition (NER) model is being engineered, capable of predicting 20 distinct entity classes. This project is designed to significantly enhance the efficiency and precision of NER in Arabic text processing by broadening the scope of entity recognition and leveraging a custom-built dataset. The ultimate goal is to achieve unprecedented levels of accuracy and performance in NER analysis, setting a new standard for Arabic text processing. * Developed a sophisticated classification model for automated tagging, extracting a substantial dataset of approximately 100,000 articles from a database utilizing SQL. Conducted extensive data cleaning and conversion processes to prepare the dataset for analysis, ensuring its suitability for machine learning and deep learning algorithms. Applied a variety of cutting-edge ML and DL algorithms to rigorously assess the model's performance. Python served as the primary programming language, enabling seamless implementation and experimentation throughout the project.

* Developed a robust pipeline for daily scraping of real estate property data, ensuring efficient storage in a database and thorough preprocessing to maintain data quality and integrity. This pipeline facilitated seamless analysis and insights generation to support decision-making processes. Additionally, a Tableau dashboard was created for monitoring and sharing analytics reports with stakeholders. * Automated risk-based models utilizing Python and AWS services, including SAM and Lambda functions, to streamline processes and enhance efficiency. This automation streamlined processes and improved efficiency, allowing for timely and accurate risk assessments. * Led the extraction of data from diverse sources such as Microsoft Dynamics and Camunda, followed by aggregation and storage in a centralized database. Employed SQL and Python for data extraction and analysis, enabling the generation of actionable insights to facilitate data-driven decision-making.

* Data Analyst for SMS-Assist, a leading US-based client, utilizing Python and SQL to perform comprehensive data analysis. Responsibilities included collaborating with the affiliate relations team to address intricate issues, detect emerging trends, and reveal valuable opportunities. * Developed an advanced Tableau dashboard that significantly transformed the research workflow for marketing professionals. This innovative tool employed web scraping techniques to extract critical data from a diverse range of stores and restaurants within a designated radius. By optimizing the data collection process, the dashboard enabled marketing specialists to precisely identify new vendors and streamline onboarding procedures. * Leveraged a range of Python libraries, including pandas, numpy, matplotlib, and plotly, to perform in-depth Exploratory Data Analysis (EDA) and execute comprehensive data cleaning tasks. These robust libraries facilitated the extraction of meaningful insights, ensuring data accuracy and reliability. * Designed and implemented a Python program that utilized latitude and longitude coordinates to identify nearby zip codes within a given city. By employing this program, I successfully generated a polygonal structure, enabling a precise visualization of the geographical distribution of zip codes in proximity.

• Patient Experience Project in which we are using patient data to predict patient overall experience whether he is satisfied or not with the overall Patient journey process. • Using data preprocessing technique prepare a cleaned data for model. All this work was done in Python language. Also created a dashboard for analyzing and visualizing data and performance of machine learning model.

• IOT Smart Management Project in which real time monitoring of rooms can be tracked through sensors. • Data Communication between sensors, sending data to Oracle IOT cloud and a Dashboard for users in which all room activity is projected in real time.

During my pursuit of a Master of Science in Computer Science, I focused my coursework primarily on the fascinating field of Artificial Intelligence (AI) and its related domains. Throughout this educational journey, I engaged in a comprehensive curriculum that blended theoretical knowledge with hands-on experience, enabling me to develop a robust understanding of AI's various facets and their applications in modern technology. Relevant Courses: Machine Learning and Pattern Recognition Deep Learning and Neural Networks Natural Language Processing Computer Vision and Image Processing Data Mining and Big Data Analytics Text Analytics

Major Courses: Image processing Softrware Engineering Internet of Things Machine Learning Database Management