✓ Concluded · Q2 2026

The Niha Builder Challenge

One dev. One problem. Ship it. A solo hackathon where you pick a real platform problem, build it end-to-end like an AI-native engineer, and ship production-ready code that’s demo-ready, mergeable, and tested.

Challenges

Streams

Days

Solo

No Teams

₹25K

Top Reward

Reward Tiers

Everyone who meets the minimum bar gets paid. No lottery, no “top 3 only.” Multiple winners per tier are expected.

🏆

Level 1 — Builder

₹25,000

Score 90–100. Production-grade. Zero known bugs. Full edge case coverage. Could merge to main today.

🥈

Level 2 — Contributor

₹15,000

Score 70–89. Solid implementation. Tests pass. Demo works. Minor gaps acceptable.

🥉

Level 3 — Starter

₹10,000

Score 50–69. Core functionality works. Has tests. Demoable. Some edge cases missing.

Below 50 → ₹0 + written feedback. Resubmissions encouraged. Speed/Quality/Teaching bonus recognition (no extra cash) for first Level 1, highest score, and standout helpers.

How It Works

Mon 4 May → Sun 10 May — 7 days, end to end.

📣

Mon 4 May

Announce & Claim

Registrations open at 12:00 AM IST. Pick one (first-come-first-served), claim it in the Teams channel: “I’m taking Challenge #N”.

📝

Tue 5 May

Setup & Plan

Fork from main, create branch challenge/{N}-{your-name}, read the brief, set up your env, ask questions.

🔨

Wed 6 – Sat 9 May

Build & Test

4 working days. Follow .dev/conventions/. Use AI tools heavily. Write tests as you go. Daily stream-owner office hours.

🚀

Mon 11 May

PR & Demo — Deadline 10:00 AM IST

Submit a single PR to main with code, tests, and CLAUDE.md updates. Record a demo video or schedule a 10-minute live walkthrough.

🎧

Mon 11 May+

Judging & Rewards

Auto-checks gate → 2 judges score per dimension → per-dimension feedback within 24h. Everyone scoring 50+ gets paid.

Judging Criteria

5 dimensions, 20 points each, 100 total. Two judges score every dimension independently — if they disagree by more than 5 points, a third judge breaks the tie.

✅

Functionality 20 pts

Does it work end-to-end, not just the happy path? Acceptance criteria met. Edge cases handled. Graceful errors.

🛠

Code Quality 20 pts

Reads like production code. Follows .dev/conventions/. Type-safe, no security issues, no dead code.

🧪

Testing 20 pts

Can someone refactor it without fear? Unit + integration tests. Edge cases covered. Coverage >60% (Level 2), >80% (Level 1).

🧱

Integration & Architecture 20 pts

Could it merge tomorrow without breaking anything? Uses platform patterns (MCP, bridge, config cascade). Clean API boundaries.

🎥

Demo & Documentation 20 pts

10-minute demo showing end-to-end flow. PR description explains what/why/how. CLAUDE.md updated where needed.

⚙️ Pre-Judging Auto-Checks (pass/fail gate)

Before any human review, your PR must pass: pytest (Python), vitest (TypeScript), ruff + eslint linting, and no hardcoded secrets. Fail the gate → the rubric isn’t scored.

View full scoring rubric, reward tiers & score calculator →

Rules & Guidelines

Solo only. No teams, no pair programming. One challenge per person — pick wisely.

Claim it in the channel. Post “I’m taking Challenge #N” in the Teams channel. First-come-first-served. If you drop out, the challenge reopens 48h later.

Branch convention. Fork from main, create branch challenge/{N}-{your-name}, e.g. challenge/03-ravi.

Follow SIE standards. Code in .dev/conventions/ (Python, TypeScript, architecture, testing, security). Linting and full test suite must pass in CI.

Use AI tools heavily — but understand every line. “The AI wrote it” isn’t an acceptable demo answer. AI-assisted is fine; AI-generated-and-not-understood is not.

Ship one PR to main. Code, tests, CLAUDE.md updates, PR description (what / how to run / how to test / edge cases / known limits), plus a 10-min demo (recorded or live).

The Niha Builder Challenge

Winners — to be announced

Rules & Guidelines