Platform Engineer — 14 years securing and running production systems across Banking, Gaming and E-Commerce.
I build control planes for fleets: provisioning, drift, observability, safe change. I like systems that are observable, automatable, and opinionated about reliability.
Currently going deep on LLM inference infrastructure.
🔭 Currently building
kri — self-hosted fleet operations platform for macOS and Linux
Bootstraps any SSH-reachable node · runs Salt states and Ansible playbooks from one interface · scores every node against a declared baseline · answers natural-language questions about fleet state
Python · FastAPI · React · SaltStack · TimescaleDB
mlx-oc — distributed LLM inference across a two-Mac-mini ring cluster
OpenAI-compatible metrics proxy recording TTFT, token rate and hallucination-risk heuristics · per-node hardware telemetry scraped every 15s · 46 alert rules across 6 groups · 12 Grafana dashboards
Python · MLX · VictoriaMetrics · OpenTelemetry · Grafana
hetu — Kubernetes cluster intelligence engine
Deterministic health scoring · four-level drill-down · Sentry-style error grouping · anomaly detection · LLM-backed root-cause analysis
Go · Next.js · Prometheus
champ — AI coding agent as a VS Code extension
Multi-agent orchestration over a shared memory system · multi-provider auto-discovery · runs fully offline against a local model · per-tool approval flow
TypeScript · Anthropic API
🛠 Tech I work with
✍️ Recent writing → hellodk.io
- 3 Root Causes, 2 Identical Errors, 1 Dead NFS Box: Consolidating a Homelab Behind SSO
- The Perils of Timezone — A Complete Visual Guide
- 27 Tests, 3 Alerts, 1 Real Bug Review Caught: Making a Drift Detector Watch Itself
📍 Bangalore, India · 📷 Photographer ·