I build AI systems that move from prototype to production.

Architect of ThreeZincAI and founder of DayZTools.ai. I design distributed inference, multimodal workflows, and GPU infrastructure built for real workloads.

View selected work Explore architecture

93%: Infrastructure cost reduction
20x: Cheaper image editing
5+: Cloud and AI providers

Selected work

Products shaped from the systems layer up.

A focused selection of platform architecture, model engineering, and founder-led product work.

01Platform architecture

ThreeZincAI

Workflow-driven AI content generation

A production platform for building, running, and scaling multimodal generation pipelines across images, shots, and video.

Graph-based workflow execution
Distributed GPU workers
Asset lineage and version tracking
Multi-provider abstraction

Explore the architecture

ThreeZincAI visual workflow canvas for a banking campaign — Workflow-driven AI content generation5+ · AI providers integrated

02Model engineering

Carousely.ai

High-volume image editing optimization

Custom Qwen Image Edit LoRAs and an optimized serving pipeline replaced expensive generic APIs while preserving edit accuracy at scale.

Purpose-built LoRA training
Curated alignment datasets
Spot-instance inference
Production model routing

Visit carousely.ai

Comparison of generic and optimized image editing model pipelines — High-volume image editing optimization20x · Lower generation cost

03Founder project

DayZTools.ai

AI-assisted game development suite

A focused toolkit that helps DayZ creators build and iterate on terrain, assets, and modding workflows with AI-assisted tools.

3D terrain workflows
Asset generation tools
Creator-first UX
Automated project setup

Visit DayZTools

DayZTools 3D terrain builder interface — AI-assisted game development suite3D · Creator workflow suite

Under the hood

Systems designed for failure, scale, and cost.

Workloads move through durable queues, provider-aware routing, and elastic GPU workers. Retries and asset state are part of the architecture, not afterthoughts.

Distributed inference pathProduction pattern

Client

Workflow engine

Global queue

GPU workers

Asset storage

Automatic retries Provider failover Asset lineage Real-time state

Expertise

From model behavior to production infrastructure.

Systems

Python
TypeScript
SQL
Distributed systems

Infrastructure

Cloudflare
Docker
WebSockets
GPU orchestration

AI engineering

Flux / SDXL
LoRA training
ControlNet
Inpainting

Platforms

dstack
RunPod
FAL / Runware
Replicate

Start a conversation

Building something technically ambitious?

I’m interested in AI infrastructure, distributed systems, and products where thoughtful engineering creates a real advantage.

Email me GitHub LinkedIn