Implementation / architecture / team leadership
Selected work
Systems I have built or helped deliver, from production data pipelines and LLM agents to GPU infrastructure and current R&D.
production
Production data pipelines
I led the team and co-developed an automated pipeline that turns unstructured, inconsistent inputs into standardized business data.
The pipeline combines AI methods with local GPU compute, automated runs, observability, and data provenance. Its storage and processing were designed to avoid redoing work unnecessarily.
production
Persistent multi-agent systems
I led the team developing a multi-agent LLM chatbot with persistent memory and presence across multiple sites.
I was responsible for technical decisions and prototypes as the team connected multiple products and agents into a production system. The role combined architecture, hands-on development, and engineering leadership.
internal use
GPU experimentation lab
I built and administered a GPU cluster for an AI engineering team's experiments.
The cluster provided compute for the team's model work. My responsibilities covered both the hardware and the systems running on it.
in development
AI software across heterogeneous hardware
At aleph0, we are developing low-level software for AI on heterogeneous hardware, especially at the edge.
The work combines hardware-specific compiled models, a single-binary runtime, and custom model optimization for compatibility and lower power use. It is in development and will be released open source when ready.
Further work includes co-developing an automated LinkedIn marketing tool with topic suggestions and tone adaptation, writing technical articles, and delivering workshops on AI, LLMs for software engineering, and safety practices.
My research page covers published work in medical imaging, phototoxicity prediction, and LLM routing.