Senior) Platform Engineer - m​/f​/d

Where Europe's enterprises adopt AI

Langdock is the AI platform used by more than 10,000 companies to give employees secure access to the leading AI models, to build and share agents and to automate repetitive workflows. We have grown past $40M ARR while remaining a small team, and we care deeply about operating efficiently across the entire company.

For many enterprises, Langdock is becoming the place where most of the net-new work is produced. As people and agents create more documents, analyses, decisions, and automations inside AI interfaces, the context and data behind that work accumulate within Langdock. This gives us the opportunity to earn a larger role in their technology stack by building a platform they choose to rely on.

Our ambition is to build that platform for European enterprises while preserving their control over data, model providers, and deployment environments. We have made meaningful progress at the application layer, but much of the foundation beneath it still needs to be built.

You can watch the Meet the engineering team video to get a feeling for how we work.

Platform Engineers own the shared capabilities that Langdock's products are built on. They treat these capabilities as internal products: defining the interfaces other teams build against and remaining accountable for their reliability, evolution, and operation over time.

The work is to turn complex capabilities such as model access, agent execution, authorization, queues, integrations, document processing, and sandboxed compute into stable, reusable building blocks. Each capability needs a clear contract and must be able to evolve without forcing coordinated changes across the rest of the product.

You own a service from design through long-term operation. That includes defining its API and failure behavior, planning migrations, implementing and deploying it, making it observable, and responding when it fails in production. When a service requires a new underlying mechanism, you work with Systems Engineers while remaining accountable for the service boundary and lifecycle.

• A durable agent runtime. The current engine runs model and tool loops for Chat, Workflows, automations, and subagents, but active execution is still owned by one process. We are separating the runtime from product‑specific state and designing durable runs that can wait for approvals or external events, recover after a process dies, and continue without repeating committed tool calls. This also includes deterministic context reconstruction, compaction, and prompt‑cache efficiency.
• The Model Gateway. We are building one inference boundary for every Langdock product, with a model‑agnostic API over managed and customer‑provided models. The work includes provider adapters, routing across accounts and regions, rate limits, bounded fallback, streaming safety, usage tracking, and a control plane.
• Shared services with explicit contracts. We are moving selected capabilities out of a Type Script monolith into independently operated services with Protobuf and gRPC contracts inside one Bazel monorepo. Our authorization service, written in Go with embedded SpiceDB, is one example: the migration uses an outbox, shadow comparisons on production traffic, and consistency tokens before it becomes authoritative.
• The integration and tool platform. Langdock already connects to external services through REST APIs, Model Context Protocol (MCP), and agent-to-agent (A2A) communication. The platform work is to make authentication, token refresh, rate limits, approvals, outbound network safety, and tool execution consistent across these paths.
• Code execution as a shared service. Langdock already runs untrusted…