Web Browser Automation

AI browser automation agent using browser-use library (Python) with locally-hosted LLM via Ollama. Stack: browser-use, Playwright, Ollama, Qwen3-VL or bu-30b-a3b-preview, Gradio UI. Build a conversational agent that takes natural-language tasks and executes them in a real browser — Amazon shopping, flight search on Momondo/Skyscanner, multi-page form filling, ticket lookups. Must include a human-in-the-loop safety gate for purchases and PII submission, action logging, and session reports. Target hardware: RTX A6000 (48GB VRAM) on Ubuntu. Must run fully offline — no cloud API dependencies. Required experience: browser-use framework, Playwright, local LLM deployment (Ollama or vLLM), vision-language models. Bonus: experience with browser-use's bu-30b-a3b-preview or Skyvern.