Hammad Aamir K.

I build complete AI applications end-to-end — LLM pipeline, API, job queue, database, billing, front end — and ship them. Not prototypes, not slide decks. 100% Job Success, Top Rated Plus, 4,000+ hours delivered. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHAT I DO ▸ AI / LLM systems — multi-stage LLM pipelines, agent orchestration (LangGraph, LangChain), Model Context Protocol (MCP) servers and clients, tool-schema and prompt design, structured and validated model output, vision models, per-run token and cost accounting, vector search (CLIP + FAISS). OpenAI and Google Gemini in production. ▸ Backends that hold up — FastAPI (async), Redis/RQ background workers, PostgreSQL with versioned migrations and stored procedures, Supabase, WebSockets and Server-Sent Events, Docker, Railway, Cloudflare, AWS/GCS, serverless GPU (Modal). ▸ Front ends that match — React 18 + TypeScript, Vite, Tailwind, real-time UI over WebSockets and SSE, and the dashboards, admin panels and onboarding flows that turn a pipeline into a product someone can actually use. ▸ Also, unusually: Blender and 3D automation — Blender Python (bpy) add-ons, procedural generation, render-farm orchestration. Very few engineers do LLM orchestration and bpy, which is how I ended up leading engineering on an AI 3D platform. ▸ The unglamorous parts that decide whether a product survives — subscription and metered billing with atomic, idempotent balance operations, webhook handling, JWT/JWKS auth, multi-tenant scoping, CI, and tests that run against real models and a real database. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ RECENT WORK • Co-founder and sole engineer of an AI marketing SaaS: a four-stage LLM pipeline behind an async FastAPI service, a Redis-backed job worker, Postgres with versioned migrations, metered credits and subscription billing built on atomic database procedures, and a React + TypeScript front end — architected, built, deployed and operated by one person. • Lead Engineer on an AI-driven 3D platform (long-running contract, still active): LLM agent systems and MCP servers driving Blender automation — natural language in, rendered scenes out, with camera, lighting, animation and render orchestration behind FastAPI services and a PySide6 desktop application. • Contract work on AI-assisted 3D product visualization: Blender add-ons on serverless GPU, LLM-driven scene setup, and CLIP/FAISS retrieval over asset libraries. • Earlier: 60+ freelance projects for 12+ international clients — numerical simulation, scientific computing and Blender production work. Independent verification matters to me: one aircraft-performance model hit RMSE under 1.5% across 142 scenarios, and several deliverables were validated against peer-reviewed publications. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ HOW I WORK I've been a technical co-founder three times, so I don't need a spec to start — give me the outcome and the constraints and I'll come back with an architecture and the trade-offs before I write anything. I work in small increments with visible progress, I say clearly when something is a bad idea, and I don't disappear mid-project. Fully remote, comfortable async, and used to long-running engagements: my current main client has been with me since 2023. I own the architecture, the AI layer and the code review on every project personally, and I have engineers who handle build-out so delivery doesn't stall. You deal with me, start to finish — one point of contact, and my name on the result. Background: BSc Mechanical Engineering, Bilkent University (3.61/4.00) — which is where I learned to model hard problems before coding them. I take on a small number of larger builds rather than many small ones, so if you have an AI product that needs to exist properly — send me the problem and I'll tell you honestly whether I'm the right person for it.