Gavin Dsouza
Machine Learning Engineer
About
I'm a Machine Learning Engineer specializing in LLMs and multimodal AI systems. Currently pursuing my M.S. in AI at UC Santa Cruz, where I'm partnering with Adobe to build an intelligent Browser Agent for the Express Platform. With 2+ years of software engineering experience, I focus on LLM fine-tuning and agentic AI applications. I'm also developing a general-purpose Browser Copilot to bring Cursor-like intelligence to web browsing.
Stuff I've Made
Protostar - Browser Copilot
Cursor for the Browser
Slug-Search
Fine-tuning LLMs using outcome-based reinforcement learning (GRPO) to strategically use search/retrieval tools.
Multi-Modal Reasoning
Training Pixtral-12B for figurative image-text matching using CoT reasoning.
Experience
Master of Science in Artificial Intelligence. Partnering with Adobe on capstone project developing a Browser Agent for Adobe Express. Research focus on agentic RAG systems, LLM fine-tuning, and multimodal AI.
Designed OCR pipeline reducing manual document processing effort by 90%. Led team building multi-task product classifier for 50,000+ fashion items. Optimized ML algorithms achieving 7x performance improvements. Built similarity search with 90% accuracy in <1 second. Started as Software Developer Intern, developing neural network regression models for product manufacturing cost estimation.
Published several Computer Vision research papers focusing on biomedical applications. Minor Specialization: Computational Intelligence. Focused on signal processing, embedded systems, and machine learning fundamentals.