Kaushik Kumaran.

Building reliable systems

I treat infrastructure as a software problem nobody has fully solved. At IBM, I build resilient systems across a global fleet and usually find clarity somewhere between a documentation deep dive and the 3am page that never comes.

GitHub LinkedIn Medium Credly Resume

Open source

Contributions

All work

NVIDIA · upstream

dcgm-exporter

GPU fleet observability work addressing a driver-availability race condition and improving MIG GPU profile reporting.

CNCF · upstream

Volcano

Hardening Kubernetes batch scheduling across admission control, scheduler plugins, and fractional GPU preemption.

CNCF · SPIFFE · upstream

Tornjak

Making identity management more portable across real-world Keycloak deployments and strengthening its CI coverage.

Hackathons

Built under a clock

All work

OpenAI Build Week

Argus

A layered Kubernetes security and resilience system combining agentic investigation, chaos testing, and verified recovery.

HackCBS 4.0

Post.io

A logistics-services product built as a full-stack TypeScript application.

Smart India Hackathon 2022

Offline OTP

A mobile approach to generating OTPs in areas without cellular connectivity.

Projects

Systems in progress

All work

Argus

AI-native Kubernetes security that turns eBPF, policy, and runtime signals into agent-assisted threat response.

Sentinel

A LangGraph multi-agent orchestrator coordinating Argus for security and Phoenix for chaos and regression testing.

Phoenix

A Kubernetes chaos and self-healing agent that injects failures, diagnoses incidents, and gates remediation behind human approval.