Описание

Preply creates learning experiences that help people find tutors, build personalized learning journeys, and stay motivated. Its platform connects over 100,000 tutors teaching more than 90 languages to learners in 180 countries.

Задачи

Build a durable agent runtime with checkpoints, retries, timeouts, resumability, and deterministic recovery where possible

Create a secure tool layer for discovery, registration, invocation, and versioning across internal APIs and external services, including MCP where useful

Own agent state, context, and memory primitives with clear lifecycle, privacy, retention, and retrieval semantics

Design identity, scoped permissions, and human controls for authorization, approvals, escalation, and handoffs

Develop guardrails for tool access, policy enforcement, argument validation, sandboxing, and action limits

Provide end-to-end observability, trace replay, and scenario-based evaluation across plans, model calls, tool calls, state transitions, and outcomes

Ship SDKs, templates, local tools, documentation, and paved-road patterns so teams can build agents without re-solving runtime infrastructure

Evaluate frameworks, workflow engines, and tool protocols

Operate the platform with clear SLOs, capacity and cost controls, incident response, and operational standards

Требования

5+ Years designing, building, and operating backend services or ML infrastructure handling production traffic

Strong Python skills

Experience building developer-facing APIs/SDKs or shared libraries

Experience building GenAI-powered or agentic systems using tool or function calling,



structured outputs, and multi-step workflows

Hands-on experience with Agentic AI, influencing the team beyond code generation, applying judgment to architecture, business logic, and security, using AI as a thought partner, and setting up quality gates and safety nets to enhance engineering workflows and propose AI-native alternatives

Experience integrating production services through APIs and events, including authentication, authorization, and secure service-to-service interaction

Experience with at least one major public cloud and its services and infrastructure; AWS is preferred

Hands-on experience with distributed tracing, system telemetry, SLOs, and incident triage

Strong communication skills

English at C1 level or higher

Willingness to join a paid on-call rotation

Будет плюсом: experience with agent frameworks or durable workflow technologies such as LangGraph, OpenAI Agents SDK, or Temporal; agent evaluation, simulation, or production feedback loops; sandboxed execution environments; delegated authorization for autonomous systems; MCP Gateway; GenAI observability and model gateways, including the boundary between a shared GenAI platform and the Agent Runtime

Условия A generous monthly allowance for lessons, a Learning & Development budget, and time off for self-development A competitive financial package with equity, leave allowance, and health insurance

Relocation package to join the Barcelona Hub

Free mental health support platforms

Access to Gympass-partnered wellness and gym centers throughout Spain

📌 ML platform engineer for agentic platform (Madrid)
🏢 Preply
📍 Madrid