Senior systems software engineer at NVIDIA on the NVIDIA Kubernetes Accelerator (NKX) team. Based in Olympia, WA.
Before NVIDIA I spent about 12 years on data and ML infrastructure at startups.
- NVIDIA/nodewright (formerly Skyhook): I'm the lead developer. I started it on NVIDIA's Linux team and took it open source. It's a Kubernetes operator for safely rolling out node OS customizations, and it runs on hundreds of NVIDIA's internal clusters.
- NVIDIA/aicr: AI Cluster Runtime. Generates validated, reproducible configurations for GPU-accelerated Kubernetes (snapshot → recipe → validate → bundle).
- Software supply chain: signing and verifying what we ship with Sigstore, SLSA provenance, SBOMs, and OpenVEX.
- 🎤 KubeCon + CloudNativeCon NA 2025: Tuning GenAI Workloads on Kubernetes: What Actually Works (and What Doesn't)? with Ishaan Sehgal
- ✍️ NVIDIA Developer Blog: Manage Kubernetes Node Fleets with NodeWright
Background
- Stream processing and IoT: real-time Wi-Fi location analytics, triangulating devices from off-the-shelf routers running custom OpenWrt builds, streamed through Apache Storm. It competed with Cisco's location services at under $1k to get started, against $50k. Also a Kafka/Spark data lake for the location data.
- Data platforms: high-volume ingestion and normalization (millions of real estate listings from 35 sources), Spark streaming, and warehousing on Snowflake, dbt, and Airflow.
- Geospatial: geocoding and market definition on Spark, OpenAddresses, libpostal, and PostGIS.
- ML infrastructure: CPU and GPU data and training pipelines on Airflow and EKS, which is how I ended up in Kubernetes.
- Distributed computing
- Data engineering
- Good abstractions
- Small, modular tools that do one thing well
Now: Go · Kubernetes · Helm · Sigstore / cosign · GitHub Actions · Terraform
Past lives: Spark · Kafka · Storm · Hadoop · Airflow · Snowflake · Scala · pandas