Python & AI Engineer — I build LLM integrations, RAG chatbots, and API automations that work in production.
📍 Lahore, Pakistan · 💼 Available for freelance work · 🗣️ English / 中文-aware tools for cross-border e-commerce
| Service | What you get | Proof |
|---|---|---|
| RAG support agents | A chatbot that answers only from your catalogue and policies, and says "I don't know" instead of guessing | ECOM-Support_agent |
| LLM integrations | Qwen / OpenRouter / LLM APIs wired into your product, including bilingual EN/中文 assistants | Bridge |
| API & webhook automation | Verified, retried, fully logged integrations between your tools (GitHub, Stripe, custom) | Relay |
| Monitoring & scheduled jobs | Self-hosted watchers and schedulers with secure auth | Watchtower |
Use case: a support bot that never makes up answers. RAG agent that answers only from a store's own catalogue and policies, and refuses to guess when the data doesn't cover a question.
- Stack: Python, FastAPI, sentence-transformers, OpenRouter
- Result: 94% correct behavior, zero hallucinations on a purpose-built eval set
Use case: customer support and product listings in two languages. Bilingual (EN/中文) support and listing assistant for cross-border e-commerce, powered by Qwen (Alibaba Cloud Model Studio) and live-tested against the real API.
- Stack: Python, FastAPI, Qwen/DashScope
Use case: never lose a webhook again. Webhook router that verifies, delivers, and automatically retries webhooks (GitHub, Stripe, or custom) with a full delivery audit trail.
- Stack: Python, FastAPI, APScheduler
Use case: get told when a page changes. Self-hosted page-change watcher. Schedules checks on any URL and detects changes by content hash, with SSRF protection and JWT auth built in.
- Stack: Python, FastAPI, APScheduler, SQLAlchemy
- Boardroom: real-time collaborative kanban board with WebSocket-synced updates and live presence. (Python, FastAPI, WebSockets, SQLAlchemy)
- ShelfWatch: fine-tuned YOLOv8 that detects empty shelf space in retail photos, served via FastAPI with a live upload dashboard. mAP@50 0.613 on a held-out validation set. (Python, PyTorch, Ultralytics)
Languages: Python, TypeScript AI/LLM: RAG pipelines, LLM integration (Qwen, OpenRouter), sentence-transformers, PyTorch, YOLOv8 Backend: FastAPI, WebSockets, PostgreSQL, SQLAlchemy, APScheduler Other: Docker, GitHub Actions, Git
I take on freelance projects in LLM integrations, RAG, and API automation.