Multi-Cloud AI Engineer β I build systems at the edge of ML infra, hardware-aware tooling, and agentic applications: compatibility checkers for kernels, GPU/HPU monitors, version control for datacenter hardware, and LLM-powered products shipped on AWS, Azure, and GCP.
- ML infrastructure & hardware β kernel/ISA compatibility tooling, Gaudi/GPU monitoring, CNN training pipelines
- Agentic AI products β multi-agent systems, RAG, voice/LLM safety tooling, built for hackathons and production alike
- Multi-cloud β AWS SageMaker, Azure ML, GCP/Gemini, deployed with real CI/CD, not just notebooks
| Project | What it does |
|---|---|
| Kernelridge | Local-first, dependency-free compatibility preflight CLI for Python β checks architecture, ISA extensions, ABI, and library versions before a deploy ships, and refuses to rubber-stamp an optimistic PASS |
| vcm-machines | Version control for datacenter hardware β commit a node's config, see what a change breaks, trace which physical units are affected, with semantic part search |
| gaudi-monitor | Local monitoring TUI, CSV logger, and Prometheus/OpenMetrics exporter for Intel Gaudi accelerators β a single <80KB binary, zero runtime dependencies |
| KinShield | Scam protection a caregiver sets up once, for a parent who does nothing β voice/script scam detection on Amazon Bedrock, live on AWS |
| all-builders (SafeShelf Global) | Checks product listings for recalls and hazards across 10 regulators before you buy β built on Gemini + Google ADK |
| wine-quality-mlops | Production ML CI/CD pipeline: GitHub Actions + Azure ML + Azure Monitor |
| threadnet-cnn-sagemaker | A CNN trained from scratch on Fashion-MNIST, entirely inside AWS SageMaker Studio |
| ds-visualizer | Interactive 3D tactile data structures visualizer with real-time algorithm stepping |


