I’m building a car-review website where the standout feature is an AI engine that tailors vehicle recommendations to each visitor. Rather than scraping browsing data or hooking into social media, the system will rely on a concise user-preferences questionnaire presented during the onboarding flow. The questionnaire must capture: • Driving style • Car brand preferences • Budget range Using those three inputs, the model should instantly rank suitable cars and surface the most relevant in-house reviews, specs, and pricing information. I need the full pipeline: data pre-processing, model selection or fine-tuning, API or microservice that returns ranked results, and a front-end component (React preferred, but open) that displays both the questionnaire and the recommendation list. Key deliverables • Responsive questionnaire UX and validation • Trained recommendation algorithm with clear reproducibility steps • Integration layer that links the algorithm to the review database and front-end • Admin interface or config file so I can update car data without redeploying • Deployment on my VPS with basic monitoring Acceptance criteria 1. A first set of recommendations appears in under one second for a typical query. 2. Accuracy: at least 80 % of test users feel the top three cars match their stated preferences. 3. Code is documented and handed over via Git, including environment setup scripts and a short README. If you have experience with Python ML libraries (e.g., scikit-learn, TensorFlow) and modern JavaScript frameworks, we’ll be on the same page. Let me know which stack you favour and how you’d approach training and evaluation, and I’ll share the current dataset so we can get moving quickly.