███╗ ███╗███████╗██╗ ██╗███████╗██╗ ██╗███████╗██╗ ██╗ ████╗ ████║██╔════╝██║ ██║██╔════╝██║ ██║╚══███╔╝██║ ██║ ██╔████╔██║█████╗ ███████║█████╗ ██║ ██║ ███╔╝ ███████║ ██║╚██╔╝██║██╔══╝ ██╔══██║██╔══╝ ██║ ██║ ███╔╝ ██╔══██║ ██║ ╚═╝ ██║███████╗██║ ██║██║ ╚██████╔╝███████╗██║ ██║ ╚═╝ ╚═╝╚══════╝╚═╝ ╚═╝╚═╝ ╚═════╝ ╚══════╝╚═╝ ╚═╝
Applied AI Engineer · Co-Founder @ Smartloop · Local AI Orchestration 🧠
20+ years shipping software. Microsoft MVP (C#). Currently deep in LLM fine-tuning, multi-agent systems, and local AI inference. Previously building infra @anaplan, dev-tools @telerik. Now building Smartloop — run AI agents on your own machine, connect tools, build custom skills, and get things done without sending your data anywhere.
- Building Smartloop — a local AI agent harness with an OpenAI-compatible API; run quantized open models on your hardware, connect MCP servers, and build custom skills, no cloud required
- Building agentic workflows to extract information and generate new content
- Fine-tuning local models and building LLM infrastructure
- Occasionally writing about LLM fine-tuning, evaluation, and agentic design patterns
- Post-Training: PEFT (LoRA/QLoRA), SFT, DPO, RLHF
- Evaluation: LLM-as-judge, offline eval harnesses, benchmark design
- Serving: OpenAI-compatible APIs, INT4/INT8 quantization, speculative decoding, distillation
- Agentic Systems: Multi-agent coordination, RAG, text-to-SQL, tool use
- Languages: Python, Go, TypeScript, C#
- 🤖 Smartloop — Local AI orchestration. Multiple small models, one smart router. Your data stays on your machine.
- 🌐: smartloop.ai
- 𝕏: @mehfuzh
- 💼: linkedin.com/in/mehfuzh
- 📧: mehfuz@gmail.com
"Think → Build → Repeat."
I have a cat named KitKat 🐱 · Puedes hablarme en español