Work

Contributing to the Kubernetes-native ML toolkit across the Python SDK and the Spark Operator.

• Designed and prototyped OpenTelemetry distributed tracing for the Kubeflow SDK — surfacing end-to-end traces, metrics, and logs across Python SDK calls and Kubernetes worker pods, closing a long-standing observability gap for ML platform teams.
• Shipped 5 PRs across the SDK and Spark Operator: a SparkCluster CRD for Spark Standalone on K8s, MultiKueue integration, ModelRegistry extensions, Trainer docs, and a stability fix.

Building ML infrastructure and a recommerce system purpose-built for ML hardware.

• Mapping aging ML hardware to specific workflows with automated resale-value calculation.
• Targeting large GPU fleets in data centers, aiming to lift ML workflow efficiency by ~50%.

Building a real-estate tokenization and mortgage solution for client deployment.

• Spanning tokenization, mortgage origination, and AI-driven valuation across MERN, Hyperledger Fabric, and FastAPI.
• Targeting the US fractional real-estate ownership market (~$4.6B) — an emerging onchain investment driver.

Built a stock-prediction system with algorithmic trading for risk management and portfolio allocation.

• Implemented Redis caching in a Unix environment, cutting API response times by 60% through efficient memory management.