Senior Backend Engineer – Remote (Brazil)

WebHosting; Brasil; Remote; BRL 180,000 - 300,000; Full time

Job summary

WebHosting in Brazil is seeking an experienced Python API engineer to design and scale async REST and WebSocket services using FastAPI, DI, and clean architecture.

You will implement multi-agent AI workflows, integrate multiple LLM providers, build RAG pipelines against vector stores, and own data models with migrations. This role emphasizes production ownership and cross-functional collaboration for robust, cost-aware AI delivery.

Responsibilities

• Design and scale async REST/WebSocket APIs in Python using FastAPI and DI.
• Implement multi-agent workflows to route work among specialized LLM agents.
• Integrate multiple LLM providers behind a provider-agnostic layer for A/B testing.
• Build Retrieval-Augmented Generation pipelines against vector stores (pgvector/Chroma).
• Design schemas and evolve data models with SQLAlchemy/SQLModel and migrations.
• Instrument services with structured logs, distributed tracing, and metrics.
• Own CI/CD, lint, type-check, test, package, and deploy for safe releases.
• Champion AI across the lifecycle with guardrails for review, provenance, and data entry.
• Move ideas from concept to production with clear architectural ownership.

Skills

Python

FastAPI

Dependency Injection

Type Hints

Async REST/WS APIs

Tools

SQLAlchemy

SQLModel

pgvector

Chroma

CI/CD

Migrations

Job description

Job Description

Job Description

What You'll Do & How You'll Make Your Mark

• Design and scale async REST/WebSocket APIs in Python (FastAPI or similar), using dependency injection, type hints, and clean, vertical-slice architecture
• Implement multi-agent workflows (sequential, concurrent, and handoff patterns) to route work among specialized LLM agents
• Integrate multiple LLM providers behind a provider-agnostic layer, enabling A/B testing and cost-aware routing rather than hard-coding a single vendor
• Build and maintain Retrieval-Augmented Generation (RAG) pipelines against vector stores such as pgvector, Chroma, or a managed vector search service
• Design schemas and evolve data models (SQLAlchemy/SQLModel-style ORM plus migrations) that support both traditional application logic and AI-driven features
• Instrument services with structured logs, distributed tracing, and cost/latency metrics, holding a real bar for p95 response time under production load
• Maintain end-to-end CI/CD ownership, lint, type-check, test, package, and deploy, so releases are boring and safe
• Champion AI across the delivery lifecycle, not just code generation, and set the guardrails the team works inside: how generated code gets reviewed, how provenance and licensing are checked, and what customer data may enter a prompt
• Move ideas from concept to production fast, with real ownership over architecture and outcomes, not just tickets