Roadmapশুরু করুন
হোম/Roadmap/Chapter 12.04
Phase 12 · Chapter 12.04

Open Source Contribution

OSS contribution = real-world code review + global resume। MLOps space-এ entry point কোথায়, কীভাবে — সব এখানে।

Why

OSS-এ contribute করার ROI

  • Top-tier engineer-এর code review পাও — free mentorship।
  • GitHub profile = global, verifiable resume।
  • Networking — maintainer-রা পরে job refer করে।
  • Deep understanding — যে tool ব্যবহার করো, internal বুঝে যাও।
  • Conference talk + blog opportunity खुले।
Targets

MLOps OSS-এ best entry projects

  • MLflow: tracking, model registry — Python, active।
  • LangChain / LlamaIndex: LLM framework, fast-moving, many "good first issue"।
  • Kubeflow / KServe: K8s + ML — Go + Python।
  • Ray: distributed compute — challenging but rewarding।
  • DVC: data versioning — Python, friendly community।
  • Triton / TorchServe: serving — C++/Python।
  • Evidently / Whylogs: monitoring — beginner-friendly।
First PR

5-step playbook

textproduction
1. Pick project   — তুমি ব্যবহার করো এমন একটা
2. Lurk 1 week    — issue tracker, Discord/Slack পড়ো
3. Find issue     — label: "good first issue" / "help wanted"
4. Comment        — "I'd like to take this, planned approach: ..."
5. Open PR        — small (<200 line), test included, link issue
PR Quality

যা reviewer খুঁজে

  • One PR = one logical change (don't bundle)।
  • Test যোগ করো — coverage না কমে।
  • CONTRIBUTING.md follow — lint, format, commit message।
  • PR description: what / why / how tested।
  • Screenshots / before-after — UI/output change হলে।
  • Reviewer feedback-এ defensive না হয়ে কৃতজ্ঞ হও।
Example

Good PR description template

markdownproduction
## What
Add INT8 quantization support to `mlflow.pytorch.log_model`.

## Why
Closes #12345. Users requested smaller artifact size for edge deploy.

## How
- New `quantize` parameter (default False) on `log_model`
- Wraps model in `torch.quantization.quantize_dynamic`
- Stores quantization config in MLmodel YAML

## Testing
- Unit: `tests/pytorch/test_quantize.py` — 6 new cases
- Integration: trained ResNet18 → log → reload → predict, accuracy delta 0.4%
- Artifact size: 44MB → 11MB

## Breaking change
None — default behavior unchanged.
Path to Maintainer

Contributor → committer → maintainer

  1. 5-10 merged PR same project-এ।
  2. Issue triage শুরু করো — duplicate close, reproduce confirm।
  3. অন্যের PR review করো — maintainer load কমায়।
  4. Design discussion-এ thoughtful comment।
  5. RFC লেখো — বড় feature propose।
  6. Maintainer-রা invite করবে — চেয়ে নয়, earn করে।
Pitfalls

OSS-এ যা beginner-রা ভুল করে

  • Drive-by typo fix scale করার চেষ্টা — maintainer বিরক্ত।
  • হঠাৎ 2000-line refactor PR — কেউ merge করবে না।
  • "why this not merged?" — চাপ দেওয়া, maintainer unpaid।
  • License + CLA না পড়ে contribute।
  • Issue ছাড়া বড় feature PR — design discussion আগে।
90-day Plan

First merged PR by day 90

  1. Day 1-14: 3 project shortlist, ব্যবহার করো।
  2. Day 15-30: codebase navigate, doc পড়ো, dev setup।
  3. Day 31-60: "good first issue" claim + PR open।
  4. Day 61-90: review iterate, merge। দ্বিতীয় PR plan।
🎉 Course Complete

তুমি যা শিখলে — full roadmap recap

  • Phase 0-2: foundation — Python, Linux, ML basics, Docker।
  • Phase 3-5: serving, CI/CD, cloud deployment।
  • Phase 6-8: orchestration, monitoring, advanced MLOps।
  • Phase 9-10: real-world AI system + architecture।
  • Phase 11: beginner → advanced production project।
  • Phase 12: career, interview, best practice, OSS।

১২ Phase, ৪৪ chapter — তুমি এখন junior থেকে senior MLOps engineer-এর full mental model পেয়ে গেছ। বাকিটা build, ship, iterate। শুভকামনা! 🚀

← Roadmap-এ ফিরুন