Custom AI ecosystem
Mout AI
Not just using AI. Understanding how it works.
From running a model to adapting Qwen and building a custom language model from scratch.
mAI Inference
A model inside a product.
Built a layer that makes an AI model usable inside an application.
Model loading, request handling, and generated output. This is an infrastructure layer, not a custom low-level inference engine.
- Inference infrastructure
RoleAI Infrastructure / Architecture / Development
mAI Sage
An existing model. Own tasks.
Adapted a Qwen-based language model for my own tasks.
Fine-tuning and adaptation experiments. The foundation is an existing Qwen model, not a model built from scratch.
- Qwen
- Fine-tuning
- Model adaptation
RoleAI Research / Fine-tuning / Experimentation
mAI CPU
Own model. From the beginning.
Developed a custom language model of approximately 300 million parameters from scratch.
Architecture, training, and inference experiments, without an existing large model as the foundation. A small experimental model for hands-on learning of LLM fundamentals.
- ~300M parameters
- Built from scratch
RoleModel Development / AI Research
Conceptual visualizations · not a running service interface
Beyond the API
The goal is to understand AI systems hands-on: how a model becomes part of a product, how to adapt it, and what it takes to build one.
From running a model to designing my own
With Mout AI, I explored three levels of working with language models: building an infrastructure layer to run them, adapting Qwen for my own tasks, and developing a custom model of approximately 300 million parameters from scratch.
This is hands-on R&D: architecture, training, and inference experiments. The aim is to build an AI system myself and understand the decisions behind a ready-made API.