Amaliyot 2: end-to-end ML system, ingestiondan servinggacha
End-to-end ML-tizim: xom ma'lumotlardan production servinggacha
O'qituvchi aniqlanmoqda
Kurs haqida
Talabalar kichik jamoalarda real datasetdan foydalanib to'liq qamrovli ML-tizim quradilar: ingestion, feature engineering, o'qitish, CI/CD, deploy va monitoring. Ushbu amaliyot MLOps va Reliable ML kurslarida olingan bilimlarni production sharoitiga imkon qadar yaqin muhitda mustahkamlaydi. Natija — SLO hujjatlashtirilgan va birinchi iteratsiya postmortemi tayyorlangan ishlaydigan servis.
Nimalarni o'rganasiz
✓2-3 kishilik jamoada rollarni taqsimlab end-to-end ML-servisni amalga oshirish
✓SLO (latency, availability, accuracy) ni hujjatlashtirish va yuklama testini o'tkazish
✓Birinchi production-iteratsiya bo'yicha root cause analysis bilan postmortem yozish
Asosiy mavzular
•ML jamoaviy ishlab chiqish: rollar, Git workflow, code review
•Great Expectations / Pandera bilan ingestion va ma'lumotlarni validatsiya qilish
•dbt yoki Feast bilan feature engineering pipeline
•Experiment tracking bilan training pipeline (MLflow/W&B)
•Versiyalash bilan REST/gRPC-servisni deploy qilish
•Yuklama testi va latency profiling
•Production'da drift monitoringi va alerting
Ushbu tavsif avtomatik tarzda yaratilgan va hali o'qituvchi tomonidan tekshirilmagan — bu muhokama uchun qoralama.