Tan N.

AI Engineer & Full-Stack Developer | LLM, RAG, Agents, SaaS Apps

I build production AI systems and modern web applications. Mostly I help startups and founders ship things that work, not just wrappers - LLM integrations, autonomous agents, SaaS products, automation that actually saves time. What I'm good at: AI engineering. RAG pipelines that don't hallucinate, agents that do useful work, models that know your domain. - RAG pipelines with hybrid search, reranking, and proper evaluation - Autonomous agents using MCP, tool calling, multi-step reasoning - Fine-tuning with PEFT/LoRA when general-purpose models aren't enough - LLM cost optimization (caching, model routing, batch processing) - Multi-model work across OpenAI, Claude, Gemini, and open-source (Llama, Mistral, Qwen) - Vector DBs: Pinecone, Qdrant, Weaviate, ChromaDB, pgvector - Custom training with PyTorch and TensorFlow when needed - Eval frameworks: RAGAS, LangSmith, custom pipelines Full-stack web development. SaaS apps end-to-end - the boring parts and the polished parts. - SaaS with auth, billing, multi-tenancy - Admin dashboards and internal tools - Real-time apps with WebSockets and SSE - Frontend: React, Next.js, TypeScript, Tailwind, shadcn/ui - Backend: Python (FastAPI), Rust (Axum), Node.js (Express, NestJS) - Databases: PostgreSQL, MongoDB, Redis, SQLite - Auth: NextAuth, Clerk, Supabase Auth, custom JWT/OAuth - Payments: Stripe, Paddle, Lightning Network for Bitcoin Automation and integration. Connecting systems that weren't designed to talk to each other. - n8n workflows (self-hosted or cloud) - Custom Python/Node scripts when n8n hits its limits - API integrations: REST, GraphQL, webhooks - ETL pipelines and data sync - Document processing: PDF parsing, OCR, structured extraction - Browser automation with Playwright and Puppeteer - Cloud platforms: Supabase, Vercel, AWS, GCP, Cloudflare Specialized backend. For when performance or infrastructure actually matters. - Rust for compute-heavy or low-latency services - Docker/Kubernetes - QEMU virtualization, embedded systems - API design that won't make your future devs hate you (REST, GraphQL, gRPC) - Performance profiling and optimization Recent work: - Pipeline processing 10,000+ queries/day, dropped API latency 15% - Local RAG system on Docker for confidential data - no third-party leaks - Bitcoin-powered cloud VM platform on Rust/Axum + Lightning Network - CRM/ticketing system handling large record volumes - Production MCP-based agents How I work: I'd rather build the simplest thing that solves your problem than the most impressive thing I could put on my portfolio. If your stack should be n8n and Supabase, I'll tell you that. If it actually needs Rust, I'll tell you that too - and I'll explain why either way. If something's blocking, you'll know within hours, not at the end of the week. If a requirement looks wrong, I'll push back with reasoning before just building it. Based in Vietnam (UTC+7), able to work with US, EU, and APAC clients. --> To start: Send me what you're building, when you need it, and your budget range. Within 24 hours I'll reply with whether I'm a fit, what I'd want to clarify, and a rough technical approach.