Natnael T.

Senior full-stack AI engineer with 5+ years of production experience. I build production LLM integrations, RAG systems and AI agents on Python and TypeScript stacks, and I use Claude Code daily to ship them fast without shipping unreviewed AI-generated code. Most of my work is owning features end to end: PostgreSQL schema, backend APIs, React and Next.js frontends, authentication and permissions, deployment and monitoring. I get productive quickly in existing codebases I didn't write, and I work independently in async teams. Selected results - elunic AG (Lead Developer): rebuilt a multi-tenant document search system with hybrid BM25 and vector retrieval, improving search performance by 92%. - QuickSync.pro: cut critical API execution time by 52% on an inventory sync platform serving 10,000+ active users. - Adify-AI: built and launched an AI video ad SaaS on the Shopify App Store, with LLM workflows designed for cost and latency. - KiwiKrave: building a real-time voice AI ordering agent with streaming speech-to-text, text-to-speech and phone call handling. - Shopify app engineering on a product that scaled successfully, plus a multi-platform food ordering system (mobile apps, kiosks, dashboards, APIs) supporting 1,000+ daily users. What I build - LLM integrations with the OpenAI and Anthropic (Claude) APIs: structured outputs, tool and function calling, prompt engineering, retries, validation, cost and latency control - RAG pipelines: document ingestion, chunking, embeddings, vector search (pgvector, Pinecone), retrieval grounded in your data - AI agents and agentic workflows, including MCP servers and tool integrations with guardrails, least-privilege access and audit logs - Evaluation datasets and regression checks, so every prompt or model change is measured instead of guessed - Full-stack SaaS: Next.js, React, TypeScript, Node.js, Python (Django, FastAPI, Flask), PostgreSQL, Supabase - REST APIs, webhooks, third-party integrations, payments and subscriptions, OAuth authentication and role-based permissions - AWS (including Bedrock), Docker, CI/CD with GitHub Actions, observability and production hardening How I use AI coding tools Claude Code is my main tool, alongside Cursor and Codex, for planning, refactoring, test generation, debugging and code review. I review, test and verify AI-generated code with the same rigor as hand-written code. I own what gets merged, not the model. How I work - Taking over a codebase: I start with a plain assessment of what is solid, what is a liability and what I would change - Small, tested, shippable increments; I document as I go - Clear written English and regular updates; overlap with US and EU hours - Long-term engagements preferred Core stack: Python, TypeScript, React, Next.js, Node.js, Django, FastAPI, PostgreSQL, Supabase, AWS, Docker, OpenAI API, Claude API, LangChain, MCP If you are running a production product and need a senior engineer to own AI features or a full-stack codebase, send an invite with a short description and I will reply with how I would approach it.