I need a clean, efficient, and well-documented pipeline of the following process USING RUNPOD API: 1. Deploy and optimize latest biggest model of Qwen3-VL on RunPod using multi-GPU setup AND appropriate settings. 2. Support uploading batches of website screenshots to the server. 3. Automatically process all images in a folder using Qwen3-VL-235B. 4. For each image, generate detailed captions + bounding boxes. 5. Save output as: 1 JSON file per image (containing full caption, detailed bounding box data, and original image filename to the JSON file). 6. Build a simple active learning / continual improvement loop: - Download output JSON files → review + edit (using image + JSON viewer) - Upload only the approved/corrected JSONs back to RunPod - Retrain / continue training a LoRA on the approved data 7. Provide simple, easy-to-use commands to repeat the full cycle (process → review → retrain). 8. Include clear documentation and instructions on how to run and update the pipeline. Provide clear, simple commands to repeat the full cycle (process → review → retrain). Include basic documentation on how to run and update the pipeline. As for the python CMD that will be the GUI/INTERFACE for running and setting the RunPod API: 1. Make sure it has the ability to load and save RunPod settings so it can auto-fill information when creating a new instance 2. Make sure the CMD in python captures and stores all information into an output text file 3. Make sure it has an overview or status and updates of each server 4. Make sure it can run multiple RunPod instances properly Deliverables: - Fully working pipeline on RunPod - Image processing script - LoRA fine-tuning / continual training script - Clear step-by-step documentation Looking for someone experienced with Qwen3-VL, RunPod multi-GPU setups, and LoRA fine-tuning. Please reply with your relevant experience and a short summary of your approach. Do not be late at delivering the project, please bid because you know how to do this project. Max project duration is 7 days or less, earlier delivery is better. Looking to do long term business with the right freelancer.