Modern One-Page AI Portfolio

I’m putting together a sleek one-page portfolio with these fixed sections: Home, About Me, Service, Project, and Contact. The design must feel unmistakably modern—clean typography, balanced white space, fluid scrolling, and pixel-perfect alignment across desktop and mobile.

Tech stack I expect you to use:
• React / Next.js for the front end
• Python / FastAPI to expose all back-end endpoints
• A lightweight AI/ML component for an in-page assistant (any well-known library is fine)
• Optional C++/C snippets if you find performance-critical pieces that truly benefit
• RESTful JSON between layers

Key features to implement
1. Single-page experience: smooth in-page navigation with deep-link support.
2. AI assistant: a simple yet functional ML model running behind FastAPI that can handle user queries (classification, recommendation—whatever minimal model you feel is quickest to demonstrate). Provide both training script and inference route so I can retrain later.
3. Telegram integration: when someone submits the Contact form, the message must be relayed to my Telegram chat. Ship an example .env showing BOT_TOKEN and CHAT_ID placeholders.
4. Production-ready look and feel: fully responsive, consistent spacing, accessible colour contrast, and zero layout shifts.
5. Clean, well-commented code plus a concise README detailing setup, env variables, build, and deployment.

Acceptance criteria
• Running `yarn build && yarn start` renders the site exactly as designed.
• `uvicorn main:app` exposes the assistant and contact endpoints without errors.
• `.env.example` holds the required Telegram keys only—nothing hard-coded.
• Submitting the Contact form sends a message that appears in Telegram instantly.
• AI assistant returns a non-trivial response generated by a real model.
• Lighthouse mobile score ≥ 90.

If you’re confident in React, FastAPI, and spinning up quick ML proofs of concept, I’d love to see your approach and timeline.