ML infrastructure
GPU clusters, distributed training, model serving and the pipelines that keep them running.
Available for collaborations
Engineer and researcher working at the intersection of artificial intelligence and computing: training and deploying models, designing ML infrastructure, and turning research into products people use.
$ whoami mario-kazela — AI engineer $ cat stack.txt python · pytorch · cuda · kubernetes · llms $ echo $LOCATION Kuala Lumpur, Malaysia▍
I'm a software engineer focused on AI computing: the models, hardware and systems that make machine learning fast, reliable and useful. I enjoy working across the stack, from GPU kernels and distributed training to the APIs and interfaces that put AI in front of users.
Lately I've been building with large language models, agents and retrieval systems, and exploring how to run them efficiently in the cloud and at the edge.
GPU clusters, distributed training, model serving and the pipelines that keep them running.
Building applications with large language models: tool use, retrieval, evaluation and safety.
Profiling and optimizing inference: quantization, batching, caching and hardware-aware design.
Deploying AI workloads on cloud platforms, containers, and resource-constrained devices.
A retrieval-augmented assistant that answers questions over private documents with cited sources.
A GPU job scheduler and dashboard for sharing compute across a small research team.
Quantized vision model running in real time on an edge device, 4× faster than the baseline.
Lead development of LLM-powered features and the infrastructure that serves them.
Built data pipelines and backend services; introduced ML model deployment workflows.
Focus on machine learning, parallel computing and systems.
Open to projects, research collaborations and conversations about AI computing.
[email protected]