Portfolio · Backend → Platform · Depth over breadth
Systems Depth Ladder — build the layer beneath your tools
Descend the stack: app → data systems → storage & consensus → the platform itself → a system at scale. Each rung is a "build-your-own-X" project that teaches a whole layer. You don't do them all — 2–3 done deeply beats a checklist. Progress saves in this browser.
⏱ Timelines assume ~6–8 focused hrs/week in your fresh pre-shift window (≈11:30 AM–6 PM), built one at a time — not on top of a night shift. Your first Go project (Tier 3) carries extra language ramp-up. The recommended 3-rung path — CareSignal + Redis + LSM/Raft ≈ 8–9 months of consistent work — fits inside your ~12-month switch window alongside DSA prep.
How to use this ladder
- Depth beats breadth. One project you can explain to the metal > five half-built repos. Finish and blog each before moving on.
- Hand-code the cores — the Raft state machine, LSM compaction, the scheduler loop. Those are the learning. AI for scaffolding & review only.
- Pick by target role. For the 40 LPA switch, ~2–3 rungs is plenty. The rest is a multi-year journey, not a sprint.
- Languages: Tiers 0–2 in TypeScript/Node (your primary); Tiers 3+ in Go (real-world alignment + your Go on-ramp).
- Write it up. A "how I built X" post per project is what recruiters find and what proves you understand it.
Prove it for ₹0 — "productionised" ≠ paid hosting
Reviewers judge how it's engineered and the evidence in your repo — most never click a live link. A sleeping free-tier app proves less than a strong README with real numbers. Zero budget changes nothing.
Free ways to run it
- Oracle Cloud "Always Free" — up to 4 ARM cores + 24 GB RAM, forever. Run a real k3s cluster here.
- kind / k3d / minikube — a real Kubernetes cluster on your laptop. The correct home for the runtime + orchestrator projects.
- docker compose up — the whole stack (app + Postgres + Redis + Prometheus + Grafana) in one command. Containerized = free.
- Fly.io / Render / Koyeb — free tiers for a small live service when you want one.
- GitHub Actions + GHCR + Pages — free CI/CD, image registry, and static hosting.
What actually convinces a reviewer — evidence in the repo
- One-command run.
docker compose up so anyone runs it in 30s — beats a hosted link.
- Killer README — architecture diagram, run steps, design decisions & tradeoffs.
- Green CI badge — test → lint → build → scan. Free, and it screams production discipline.
- The numbers. Paste k6 results (req/s, p99), Grafana dashboard screenshots, and a GIF of it running — including a node dying and self-healing.
- Tests + observability wired up — running locally is completely fine.
- Live demo when asked — spin up a free tier just for the day, or tunnel localhost with
cloudflared / ngrok during the call.