Senior Python / AI Agent Engineer

Зарплата: 1000–2000 USD (на руки)

Senior Python / AI Agent Engineer

Biz AI asosidagi mahsulotlar, intelligent agentlar, RAG tizimlari va biznes jarayonlarini avtomatlashtirish bilan ishlay oladigan tajribali Senior Python / AI Agent Engineer qidirmoqdamiz.

Nomzod faqat LLM API'larini chaqirishni emas, balki production darajasida ishlaydigan AI agent arxitekturasini loyihalash, ishlab chiqish, test qilish va optimallashtirishni bilishi kerak.

Asosiy talablar

• Python bo‘yicha kamida 4–5+ yil professional tajriba

• Python'ni advanced darajada bilish:

  • AsyncIO

  • multiprocessing / threading

  • decorators

  • generators

  • context managers

  • type hints

  • Pydantic

  • dependency management

  • clean architecture

  • SOLID va design patterns

• FastAPI yoki Django bilan production backend yaratish tajribasi

• REST API va WebSocket bilan ishlash

• PostgreSQL, Redis va SQL bo‘yicha kuchli bilim

• Docker va Linux bilan erkin ishlash

• Git/GitHub workflow'larini yaxshi bilish

AI va LLM bo‘yicha talablar

Nomzod zamonaviy Large Language Model ekotizimini chuqur tushunishi kerak.

Quyidagilar bilan ishlash tajribasi talab qilinadi:

• OpenAI API

• Anthropic Claude API

• Google Gemini

• Open-source LLM'lar

• Hugging Face

• Ollama / vLLM yoki shu kabi inference yechimlari

Quyidagi tushunchalarni yaxshi bilishi kerak:

• system/user/assistant message architecture

• prompt engineering

• structured outputs

• JSON schema

• tool/function calling

• context window management

• token optimization

• streaming responses

• embeddings

• semantic similarity

• reranking

• hallucinationlarni kamaytirish

• LLM evaluation

• guardrails

• model fallback va routing

AI Agentlar bo‘yicha kuchli bilim

Bu pozitsiyaning eng muhim talabi — AI Agent systems.

Nomzod quyidagilarni amaliy darajada bilishi kerak:

• Tool Calling

• Function Calling

• Agent Loop

• ReAct

• Planning

• Reflection

• Agent Memory

• Short-term va Long-term Memory

• State Management

• Human-in-the-loop

• Multi-agent systems

• Agent orchestration

• Agent handoff

• Background workflows

• Retry/fallback strategiyalari

• Agent permissions va security

• Agent observability

Agent foydalanuvchi so‘rovini tushunib, mustaqil ravishda kerakli tool'larni tanlashi va bir nechta bosqichli vazifalarni bajara oladigan tizimlarni qurish tajribasi bo‘lishi kerak.

Masalan:

User → AI Agent → Planning → Search/Database/API/Tool → Reasoning → Action → Validation → Final Response

AI Agent Frameworklar

Quyidagilardan kamida bir nechtasi bilan real loyiha qilgan bo‘lishi afzal:

• LangChain

• LangGraph

• OpenAI Agents SDK

• CrewAI

• AutoGen

• LlamaIndex

• PydanticAI

• Semantic Kernel

Frameworkdan foydalanishning o‘zi yetarli emas. Nomzod framework ortidagi agent architecture va state-machine prinsiplarini tushunishi kerak.

MCP — Model Context Protocol

MCP bilan ishlash tajribasi katta ustunlik hisoblanadi.

Nomzod quyidagilarni tushunishi kerak:

• MCP Server

• MCP Client

• Tools

• Resources

• Prompts

• Authentication

• External system integration

AI agentlarni quyidagi tizimlar bilan bog‘lay olish:

• Database

• CRM

• ERP

• Telegram

• Email

• Google services

• GitHub

• Internal API

• boshqa biznes tizimlari

RAG — Retrieval-Augmented Generation

Production darajadagi RAG sistemalarini yaratish tajribasi talab qilinadi.

Quyidagilarni bilishi kerak:

• document ingestion

• chunking strategiyalari

• embedding generation

• metadata filtering

• vector search

• hybrid search

• reranking

• query rewriting

• contextual retrieval

• citations/source attribution

• RAG evaluation

Vector database'lar:

• pgvector

• Qdrant

• Pinecone

• Weaviate

• Milvus

• Chroma

Ulardan kamida bittasi bilan production tajribasi bo‘lishi kerak.

AI Memory Architecture

AI agent memory tizimlarini yaratishni tushunishi kerak:

• Conversation Memory

• Working Memory

• Long-term Memory

• User Memory

• Episodic Memory

• Semantic Memory

Embedding asosidagi memory retrieval va foydalanuvchi kontekstini boshqarish tajribasi katta ustunlik.

Integratsiyalar

AI agentlarni tashqi servislar bilan bog‘lash tajribasi:

• Telegram Bot API

• Gmail / Email

• Google Drive

• Google Calendar

• Slack

• GitHub

• CRM / ERP

• REST API

• Webhooks

OAuth 2.0 va API authentication mexanizmlarini tushunishi kerak.

AI Workflow Automation

Quyidagi platformalar bilan tajriba ustunlik beradi:

• n8n

• Temporal

• Celery

• RabbitMQ

• Kafka

AI agent + workflow automation arxitekturasini ishlab chiqa olishi kerak.

Masalan:

Email keladi → Agent analiz qiladi → hujjatlarni o‘qiydi → ma’lumotlar bazasidan tekshiradi → qaror chiqaradi → kerakli API'ni chaqiradi → natijani foydalanuvchiga yuboradi.

Nomzoddan kutadigan amaliy ko‘nikma

Masalan, quyidagi vazifa berilganda mustaqil arxitektura qura olishi kerak:

“Foydalanuvchi AI agentga topshiriq beradi. Agent PostgreSQL bazadan ma’lumot oladi, internet yoki knowledge base'dan kerakli ma’lumotlarni qidiradi, hujjatlarni o‘qiydi, kerak bo‘lsa boshqa agentga vazifa beradi, natijani tekshiradi va foydalanuvchiga manbalar bilan javob beradi.”

Nomzod bu jarayon uchun:

• architecture

• database

• agent state

• tools

• memory

• RAG

• queue

• observability

• security

• deployment

qatlamlarini mustaqil loyihalay olishi kerak.

Soft Skills

• Muammoni mustaqil tahlil qila olish

• Faqat task bajaruvchi emas, yechim taklif qila olish

• Texnik qarorlarni asoslab bera olish

• Code Review qilish

• Junior/Middle dasturchilarga mentorlik qilish

• Documentation yozish

• Product va biznes talablardan texnik yechim chiqarish

Ingliz tili

Kamida B1/B2 daraja.

AI va software engineering bo‘yicha technical documentation'larni mustaqil o‘qib, tushuna olishi shart.