I build AI systems that don't break in production and infrastructure that doesn't bankrupt you.

Currently shipping multi-agent systems, on-prem LLMs, and RAG pipelines at SurePayroll. Previously ran TechMade (AI/full-stack agency, 2021–2025) with clients including VerySweet and Route.

Chicago · yazan.io · hi@yazan.io

• AI systems that fail in predictable, recoverable ways — specification design, evaluation harnesses, failure pattern recognition
• Local LLM deployments — fine-tuning Qwen and similar on Apple Silicon, on-prem inference, cost-per-token economics
• Civic tech — property ownership transparency, accountability tools

Languages: TypeScript · Python · Go · Kotlin Frontend: React · Next.js · Svelte · Vue Backend: Node · NestJS · FastAPI · Django · GraphQL AI/ML: LLM fine-tuning · RAG · Agent orchestration · Vector DBs · MLX · Qwen / DeepSeek Data: PostgreSQL · MongoDB · Redis Infra: Docker · Kubernetes · GCP · AWS · Proxmox

• OpenSpar — Open-source comparative AI analysis platform
• Who Owns What — Chicago — Chicago property ownership data
• Jail.app — Whisper model rewrite for serverless infrastructure

Open to senior IC and staff roles in AI engineering, full-stack, and platform.