Senior AI Engineer

Описание вакансии

Location: On-site in Tashkent

Role Overview

— Lead development of our AI-native productivity platform spanning a FastAPI service layer, Telegram bot interface, and LangGraph, LangChain, DeepAgents-powered autonomous agents.

— Own end-to-end Python delivery: async APIs, Celery pipelines, data models, and integrations with LLM tooling, Redis, and PostgreSQL.

— Champion reliability (pytest, coverage, mypy, ruff,) and observability (Loguru JSON logs, Sentry, Prometheus, LangSmith) in a Docker- first workflow.

What You’ll Build

— Async FastAPI endpoints and background orchestrations that drive founder and team-focused automations.

— Celery workers/beat schedules coordinating reminders, task dispatchers, and memory compression jobs.

— SQLAlchemy models, migrations, and data access around PostgreSQL + Redis.

— Memory/agent services using LangGraph, LangChain, Mem0, and Qdrant vector search.

— Telegram bot flows via python-telegram-bot, integrating with our AI agents and Redis conversation store.

Must-Have Skills

— 6+ years professional Python with deep async expertise (FastAPI/Starlette, asyncio patterns, dependency injection).

— Production Celery experience: queue design, scheduling, observability across Redis backends.

— Strong SQLAlchemy/Alembic skills; comfortable tuning PostgreSQL schemas and async session lifecycle.

— Building and hardening API ecosystems (AuthN/Z with python-jose/passlib, rate limiting via SlowAPI).

— Hands-on with Redis (caching, broker, pub/sub) and message-driven architectures.

— AI/LLM integration experience—LangChain, LangGraph, OpenAI/Anthropic/Gemini/Groq clients, or custom agent workflows.

— Docker/Docker Compose fluency; capable of running multi-service stacks locally and in CI/CD.

— Testing discipline: pytest (including pytest-asyncio), fixtures,mocking, coverage, property-based or contract tests.

— Code quality guardrails: Black, Ruff, mypy, iSort, pre-commit pipelines.

— Observability mindset with structured logging (Loguru), Sentry tracing, LangSmith, and Grafana metrics.

Nice to Have

— Production experience with LangGraph, LangChain or similar multi-agent orchestration frameworks.

— Vector databases (Qdrant, Pinecone) and retrieval-augmented memory systems (Mem0, LLMLingua compression).

— Telegram Bot API or other conversational UX tooling.

— Background in task automation for knowledge workers or workflow SaaS products.

How We Work

— Python 3.13 codebase, type hinted, enforcing conventional commits and PR hygiene.

— Docker Compose stack (Postgres, Redis, Qdrant, Flower, Celery worker/beat, API, Bot).