At Trident Group, I built and trained an XGBoost and Random Forest classification model for towel quality inspection. Using features engineered from a manually labeled image dataset, it reached 85% held-out test accuracy and replaced a six-person manual process.
I built an AI News Agent with a two-step agentic RAG pipeline using Gemini 2.5 Flash and Tavily search. It grounds each claim in a clickable citation, drops uncited claims, and includes a pytest suite for mocked LLM and search calls.
For the Google Cloud GenAI Hackathon, I built five applications with Vertex AI, Gemini, and Imagen, earning a Top 10 finish among 185,000 teams in Gold League, Phase 1. I also built Logbook, a Next.js and Supabase platform with Gemini API integration, and a LinkedIn posting agent with a human approval step.

