I build custom AI bots, automation pipelines, trained models, generative design systems, and production deployments — from the first dataset to the interface people actually use.
PixelGPT combines dataset engineering, model training, browser inference, and interface design in one public project.
PixelGPT is a fully local model that generates native 24×24, five-color pixel art from text—no cloud API, no upscaling, and no image credits.
Choose a service to see the actual deliverables, architecture, and type of outcome I can build.
Custom assistants for web, Telegram, Discord, or internal tools. I connect models to your data, APIs, approval rules, analytics, and business workflows.
These diagrams are not decoration. Switch pipelines, inspect a stage, or let them run to see how I structure production AI work.
A production assistant must retrieve the right context, use tools safely, and leave a trace that can be evaluated and improved.
Open each case to see the role, technical scope, and value delivered.
Prepared data, ran experiments, fine-tuned models, packaged inference, deployed the system, and iterated quality from product feedback.
Multichannel integrations, RAG and agents, analytics dashboards, QA loops, documentation, and team handover.
Custom creative models, real-time browser experiences, animated assets, and blockchain-ready production pipelines.
Built a foundation in high-quality visual production, rapid iteration, client communication, and the careful translation of design intent into repeatable systems.
This lightweight planner recommends a starting scope. It is not an automatic quote — it helps us begin with the right conversation.
End-to-end AI ecosystems, custom model integration, RAG and agent workflows, analytics dashboards, design automation, documentation, training, and post-launch support.
Trained, fine-tuned, packaged, and deployed custom models for a production application; improved quality through testing and product feedback.
Visual production, design systems, high-detail quality control, rapid iteration, and client-oriented delivery in a demanding creative environment.
Built custom creative pipelines and interactive assets; launched more than 30 collections on Tezos with a combined value above $50K.
Open-source LLM and vision fine-tuning, multimodal systems, generative architectures, quantization, pruning, ONNX/TFLite, and cloud/edge inference.
Send the problem, current workflow, or product idea. I can help define the architecture, build the system, and take it through deployment.