Most of my serious work is in Node.js, Python, and Golang. I like to automate stuff that annoys me and design systems that are boring in the right ways: predictable, and easy to fix when they break.
I work at ToolJet as a backend engineer, on our open-source low-code platform.
Before that I had a short gig at 5Paisa where I worked on their new stock-broking mobile app.
I started my career at Nudge building analytics and server-driven UI SDKs.
Outside of work and terminals: I rice Linux desktops (I use Arch, btw), read distributed systems papers, go on hikes, click photos.
I write about things that didn't click until they did β scripts, systems, and sometimes even life (but mostly scripts and systems).
The site you're on. A custom event-analytics SDK and pipeline I built instead of dropping in Google Analytics: session and visitor identity, real dwell tracking, and a live dashboard, all hand-rolled.
- Next.js
- React
- TypeScript
- PostgreSQL
- Redis
- Cloudflare
A personal health-data aggregator. An iPhone Shortcut pushes Apple HealthKit data (workouts, heart rate, VOβ max) to an API that stores it idempotently and computes the training-load metrics coaches actually use.
- Bun
- Hono
- TypeScript
- PostgreSQL
- Redis
- Cloudflare R2
One MCP server that gives an AI typed, live access to both my portfolio analytics and my health data, behind a single bearer token, so I can ask cross-domain questions a dashboard can't answer.
- Bun
- Hono
- TypeScript
- MCP SDK
- Zod
Max RPS
42k
β single node
p99 latency
4ms
under load
p50 latency
0.8ms
β cached
Uptime
99.97%
30d rolling
High-throughput event routing engine built in Golang. Persistent queues, at-least-once delivery, and a declarative rule DSL for fan-out logic.
- Golang
- NATS
- Redis
- Postgres
- gRPC
A lightweight media processing pipeline. Go API server handles uploads and queues jobs via Redis; Python workers process images and videos into optimized variants, storing results in cloud storage.
- Golang
- Python
- Redis
- PostgreSQL
- GCS
- Docker
- Kubernetes
Query rate
~0.7qps
5m rate, sample window
Cache hit rate
~20%
vs. forwarded
Ad-block rate
~0blocks/5m
no ad traffic in this sample
Upstream p99
~450ms
peak ~780ms
An ESP32-S3-based edge DNS appliance: authoritative for a runtime-managed record set, forwarding + caching for everything else, ad-block sinkholing, signed self-updating OTA, and Prometheus/Grafana observability β all on a single microcontroller.
- C++
- ESP-IDF
- FreeRTOS
- Prometheus
- Grafana