Projects
Case studies — problem, approach, and outcome — across research code and applied systems.
Halving pipeline turnaround with profiling + right-sized Kubernetes
Cut end-to-end turnaround of critical data pipelines by roughly half via targeted profiling, right-sized Kubernetes resources, and workload-tailored container images — validated with load testing.
Real-time pipeline-health monitoring dashboard
Designed and shipped a production dashboard for real-time data-pipeline (Airflow DAG) health, with authentication and access controls — adopted as a team's daily operational tool.
Standardized CI/CD across a multi-repo codebase
Rolled out reusable CI/CD templates (automated test, build, publish) to full adoption across a multi-repository codebase, cutting deployment failures and manual steps.
Secure, privacy-preserving LLM workflows
Privacy-safe assistants and guardrails that let teams use LLMs over sensitive internal material without data leaving the trust boundary.
Cross-region Kubernetes connectivity with an eBPF cluster mesh
Connected Kubernetes clusters across regions with an eBPF cluster mesh — global service discovery and direct pod-to-pod connectivity, no gateways, proxies, or VPNs.
GRUB: graph-informed pure-exploration bandits
Open-source bandit algorithms that use a similarity graph over arms to find the best (or good-enough) option with far fewer samples than standard UCB.