Research Resident at Machine Learning Institute

Led end-to-end ML solution development, including data acquisition, preprocessing, model development, and evaluation. Built a Transformer encoder from scratch in PyTorch for image classification, implementing self-attention and positional encoding. Designed and built an image caption generation system leveraging a fine-tuned OpenAI Vision Transformer encoder and a custom PyTorch decoder to generate descriptive captions for Flickr30k. Utilized Python libraries (Pandas, NumPy, SQL/PostgreSQL) for data manipulation and querying, and applied advanced ML techniques including parameter-efficient fine-tuning (PEFT), knowledge distillation, soft prompting, and experiment tracking (Weights & Biases) to optimize performance and efficiency.

Investment Analyst at Patron Capital

Led development and implementation of quantitative models, including financial forecasts, sensitivity analyses, and multi-scenario simulations, to evaluate the viability and risk of complex investment proposals. Analyzed KPIs such as revenue drivers, customer acquisition/retention costs, and churn probability, validating business plan assumptions and identifying risks and opportunities, recommending data-driven strategic adjustments to optimize performance against target metrics. Owned the end-to-end process of translating datasets (financial models, market research, risk assessments) into clear, data-driven narratives and recommendations for investment committees.