01 / cover · operator dossier
GOLU//Chitranshu Dhakad · Backend & Systems Developer
Deployment engines, databases, and networking tools — built from scratch.
operator active · jaipur, india · open to backend / systems roles — 26.91°n 75.78°e
I started at 17, stopped once, then came back determined to learn the layers most tutorials hide. Now I build protocols and infrastructure from scratch — because understanding failure modes is how I learn. DDIA is my bedside reading; my daily driver is a 150-minute DSA protocol followed by six hours of systems engineering.
no self-rated percentages — every layer links to a repo that uses it
live feedsource: github rest api + contribution grid · synced 2026-09-01 00:19Z · cache 1h
Initial attempt at coding at 17. Failed — lack of focus and hardware.
— ORIGINAcquired own machine and internet. Committed to the “Game Over” mindset.
— RESTARTPivoted from tutorial-MERN to backend plumbing. Raw net sockets, no frameworks.
— PIVOTLaunched the flight plan: zero → 1Cr-level systems SWE by Jan 2027. Public daily grind.
— LAUNCHBuilt video-streamer engine: range requests, pre-transcoded HLS, O(1) L2 LRU cache, throttling.
— PROJECTStress-tested concurrency in two runtimes — 1M Node scan tasks and 12M Go goroutines. macOS unresponsive. Postmortems logged below.
— STRESS TESTBuilding AeroCloud — self-hosted PaaS with dockerode, SQLite routing DB and dynamic subdomain proxy. DSA daily in TS + Go.
— ACTIVERaft consensus DB, Docker-from-scratch (cgroups/namespaces), then SaaS launch + elite SWE hunting.
— PHASE 3/4▸ rendered from lib/data.ts — add a project object and a card appears. no markup required.
High-concurrency seekable HLS streaming server. AES-128 segment encryption, GOP-aligned FFmpeg transcoding pipelines, and an O(1) L2 in-memory LRU cache built from a Map + doubly-linked list.
Custom version control system in TypeScript. Content-addressable object storage and directory hierarchies serialized into Merkle trees for immutable state.
In-memory TCP socket cache mimicking Redis. Custom RESP parser that survives TCP stream fragmentation, with command pipelining and key eviction.
Self-hosted developer PaaS engine — programmatic Docker builds via dockerode, SQLite routing DB, ephemeral port bindings and dynamic subdomain proxying. The current grind.
Concurrent TCP port scanner + banner grabber in Go — worker pools, read deadlines, signature-based service identification. Survived a 12M-goroutine stress test (see postmortem).
AI-powered documentation synchronizer published on npm. Watches git diff events and uses the Gemini API to auto-update system READMEs.
Push code, get a subdomain. AeroCloud automates local build scripts, streams live build logs over the Docker socket, assigns ephemeral ports, persists routes in SQLite, and reverse-proxies *.aerocloud.local to the right container. A Render/Dokku clone — built to own the whole request path.
Two runtimes, same lesson. Deliberate stress tests of my own concurrency models — each one ended with an unresponsive machine and a precise understanding of why.
Most of the 12M goroutines never scanned a port — there were only 65,535 jobs. The 100-slot channel applied backpressure correctly; the mistake was unbounded producer creation.
evidence · commit 3a3e31c ↗Same architecture, same lesson: a bounded 100-slot queue in front of a slow console consumer left a million pending scan tasks waiting — the queue didn't fail, it did its job and said 'slow down'.
Seeking backend / systems engineering opportunities, open-source collaboration, and hard problems involving networks, protocols and high-performance runtimes. If you have one — open an uplink.
uplink ready — 09/09 sections transmitted · response latency < 24h
chitranshudhakad.dev@gmail.com