Platform Engineer - m/f/d

Job description

Help Us Change the Way the World Works Build something that matters. Langdock exists to change the way the world works, bridging the gap between what technology can do and what people actually do with it. We bring all leading AI models into one secure, model-agnostic platform and make them usable across entire organizations. Over 10,000 companies use our platform every day, from fast-growing startups to some of Europe's largest enterprises. Their employees open Langdock to draft strategies, analyze documents, or automate workflows - helping them to work smarter, think more creatively, and reach their full potential. About the Role Platform Engineers at Langdock work on shared backend systems that many product features depend on. This includes the AI engine, queues, document processing, integrations, code execution, authentication, and billing. The job is to make these systems reliable and understandable enough that other engineers can build on them. That means designing clear APIs, choosing the right data models, handling failure cases, writing tests for important invariants, adding useful observability, and keeping abstractions simple enough to maintain. Typical problems include deciding which parts of a long conversation to keep in context, retrying background jobs without running the same side effect twice, handling model-provider failures during streaming responses, refreshing integration tokens before they expire, and enforcing tenant boundaries in shared services. What You Will Do Examples of the kind of systems Platform Engineers own: The AI engine at the core of the platform. It handles the prompts users send through Langdock and abstracts over providers such as OpenAI, Anthropic, Google, Azure, Bedrock, Mistral, and open-source models. The work includes prompt caching, routing, failover across model deployments, and normalizing provider-specific behavior behind a stable internal interface. The workflow runtime. Workflows need to execute reliably across agent steps, conditions, loops, structured-output extraction, human-in-the-loop pauses, and actions across hundreds of integrations. The platform work is about making that execution model predictable, observable, and safe to extend. Context-window optimization. Long conversations, uploaded files, and tool calls need to stay correct while model costs stay under control. This is one of the most cost-se…