Worked as an independent contractor for leading AI research organizations to train the latest large language models (LLMs). My work focused on developing the models' ability to answer complex and specialized prompts on topics related to Mathematics, Physics, Astrophysics and Cosmology.

• Deployed and fine-tuned large language models (LLMs) such as Llama 3 & Gemma on Google Cloud, integrating them with SQL databases and the Front-End for seamless user interactions.• Developed interactive Tableau dashboards to display healthcare professionals' locations on maps, linking them with the Front-End for real-time data visualization.

• Explored multiple AI tools to generate 3D facial models from 2D images, supporting healthcare applications in dermatology and plastic surgery, while working remotely for a South Korea-based AI startup.• Conducted a comparative analysis of AI models, achieving an identity similarity score of 0.78 withZero123++, and proposed a hybrid solution incorporating Deep3DFaceRecon and other tools to enhancetexture mapping, multi-view consistency, and 3D editing capabilities for healthcare applications.

In my role as an AI Trainer at Outlier, I specialize in assessing and enhancing the quality of AI-generated code. My responsibilities include:• Evaluating AI Output: I meticulously review AI-generated code to ensure it adheres to high standards of accuracy, efficiency, and functionality. This involves a thorough analysis of code logic, performance, and compliance with current industry practices.• Development and Problem Solving: I design complex coding and data science challenges that require an in-depth understanding of algorithms and real-world applications. These challenges are used to evaluate the problem-solving capabilities of our language models, helping to refine their performance.• Creating Human-Readable Summaries: An essential part of my role is to demystify AI decisions for broader audiences. I craft detailed summaries that clarify the AI's coding choices, highlight areas for improvement, and provide feedback that informs further training of our language models.• Testing and Quality Assurance: I develop comprehensive test suites to ensure that code generated by our language models performs reliably across various scenarios. My work spans unit testing, integration testing, and end-to-end system testing, ensuring robustness and efficacy in real-world applications.

• Led a 3-year data science project in collaboration with an international, cross-functional team. Built pipelines using MySQL and BASH to construct, clean, and filter a 10 TB dataset.• Reduced software runtime by 8x with code optimisation and OpenMP parallelization, saving over $50,000 in costs and reducing the project’s carbon footprint.• Developed software using Python and C++ to run simulations, perform end-to-end data analysis, doubling analysis precision.• Created dynamic dashboards and visualisations to uncover complex data patterns using Power BI andPython (Matplotlib, seaborn).

• Led 3 data-intensive projects, creating and optimising software for data analysis using Python and R,demonstrating strong programming and software development skills.• Developed supervised deep learning models (CNN, Decision Trees) for regression, achieving a 2% errormargin in estimating distances of millions of astronomical objects from imaging data.• Conducted computational astrophysics research, gaining advanced physics and statistics skills, presentingfindings at 8 international conferences and via 5 peer-reviewed technical reports.