At Tenasol, I designed and run the backend for a document-AI pipeline that processes up to 100 million pages per day. It scales from zero to more than 1,000 production VMs for one-week customer runs, then back down between runs.
I built the deployment wrappers and Terraform automation behind consistent fleet rollouts, along with fault-tolerance features such as retries, backpressure, and idempotent reprocessing. I also built the CI/CD pipeline from scratch, cutting deployment time from days to minutes, and created a custom monitoring system with Grafana dashboards.
Earlier, at FujiFilm Dimatix, I designed a compute cluster and wrote a web-based parallelizer that helped cut the design cycle time tenfold. I also build AI-assisted engineering workflows, including Project Jarvis, a distributed multi-agent LLM orchestrator running on Docker and Kubernetes.
