Working range
- Model behavior and task fit
- Retrieval quality and answer grounding
- Local inference and hardware limits
- Computer vision and QR workflows
- Containerized demos and protected endpoints
- Clear notes for technical and non-technical readers
Profile
I work across AI systems, automation, infrastructure, and applied research, with an emphasis on building small concrete systems and learning from how they behave under real constraints.
Local model evaluation, RAG retrieval quality, prompt routing, and the practical boundary between automation and human judgment. I am also returning to computer vision, especially modern object-detection pipelines and NVIDIA Parallel Box Decoding, which is still work in progress.
For project discussion or professional contact, use LinkedIn or email.