Building AI Infrastructure, Distributed Inference Systems, and Workflow Orchestration Platforms.
AI Systems Engineer with experience architecting production AI systems, multimodal generation platforms, distributed GPU infrastructure, and workflow execution engines.
ThreeZincAI
Workflow-driven AI Content Generation Platform
Built ThreeZincAI to help creative teams build, run and scale complex AI generation pipelines across images, shots and videos.
- Visual workflow builder
- Multi-stage generation pipelines
- Asset lineage & version tracking
- Distributed execution on GPU workers
- Provider abstraction (FAL, Runware, Replicate, etc.)
- History viewer
- Canvas sharing
- Real-time collaboration
Carousely.ai
High-Volume Image Editing Optimization
Fine-tuned custom Qwen Image Edit LoRAs to optimize high-volume image editing pipelines. This architectural optimization reduced overall generation costs by 20x, saving hundreds of thousands of dollars annually while maintaining pixel-level text-to-image edit accuracy.
- Custom Qwen Image Edit LoRA fine-tuning
- 20x cost reduction compared to generic VLM APIs
- Saved hundreds of thousands of dollars annually
- Sub-second image editing latency
- Custom dataset curation & alignment
DayZTools.ai
AI-Powered Game Development & Modding Suite
Founded and architected DayZTools.ai, a professional modding platform that helps game developers create 3D assets, pack PBO archives, and generate signed keys. Powered by custom dataset curation, WebGL preview frameworks, and generative AI nodes.
- WebGL 3D Asset Builder & preview canvas
- Instant PBO Modpacker with ZIP auto-binarization
- Secure .bikey / .biprivatekey mod-signing suite
- Generative loading screen & logo engines
- Serving over 1,000+ active game developers and modders
Distributed Inference Platform
GPU Infrastructure & Cost Optimization
Automatically spins up spot instances instead of expensive on-demand instances. By leveraging the dstack platform, the system dynamically orchestrates and provisions the cheapest spot instances available across continents from multiple platforms like RunPod, Vast.ai, AWS, Google Cloud, and Azure.
- Spot instance orchestration
- Dynamic workload placement
- Multi-cloud failover
- Auto-scaling GPU workers
- Idle termination & cleanup
- Intelligent retry & batching
Generative AI Pipelines
Custom Model Workflows
Specialized in configuring end-to-end generative pipelines—including Stable Diffusion XL, Flux, custom ControlNets, and IP-Adapter layers. Experienced in setting up high-precision models for identity preservation, character consistency, inpainting, and product photography enhancement.
Custom Node Workflows
Cloth Consistency Workflow
Maintains absolute clothing and outfit consistency across multiple generated frames, characters, and angles. Crucial for storytelling and visual continuity.
