Senior Backend / AI Engineer (2 roles) · Swisper

Summary: Architect the agent platform behind Europe's most ambitious AI assistant. About Swisper Swisper is a Swiss-hosted, privacy-first AI assistant that goes beyond chat to real execution. Hybrid LLM orchestration, contract-driven workflows, and a strict zero-data-leak promise put European data sovereignty at the core of everything we do. Swisper is built by Fintama, an established fintech consultancy already trusted by Barclays, HSBC, Deutsche Börse and others. The two backend engineers we hire will own the agent architecture that makes Swisper actually work. Why this role stands out Real agent systems, in production. Not demos. Multi-agent orchestration, tool use, memory, evals — running for real users under real privacy constraints. Architectural ownership. You're not implementing someone else's design. You decide how agents are composed, how state flows, how tools are contracted, how we evaluate. Mission with teeth. Swiss hosting, zero-data-leak guarantees, regulated-industry customers. The constraints are the interesting part. Lean senior team, weekly releases, direct founder access. Path to permanent for the right person. What you'll do Design and build the agent orchestration layer powering Swisper — planning, tool-calling, memory, multi-agent handoffs Build the developer tooling and SDKs that let us (and eventually third parties) ship new agents and tools quickly and safely Own the prompt engineering and evaluation stack — versioned prompts, regression evals, observability into agent behaviour Integrate hybrid LLM backends — frontier APIs plus self-hosted models (Ollama, local inference) under strict data-locality rules Work with vector stores and retrieval pipelines for grounded, auditable agent behaviour Partner with frontend on streaming APIs, WebSockets, and tool-trace interfaces Set the bar for code quality, testing, and architectural review Required Expert Python — async, typing, modern tooling (uv/poetry, ruff, pydantic) Strong architectural judgment. You can defend why a system is shaped the way it is, and you've earned that opinion the hard way Deep experience with agent frameworks and patterns — LangGraph, LlamaIndex, custom orchestrators, ReAct/plan-execute/multi-agent patterns Serious prompt engineering skills — structured outputs, tool-calling, eval-driven iteration, not vibes Production experience with LLM APIs (Anthropic, OpenAI, Google) and self-hosted inference Vector databases and retrieval — pgvector, Qdrant, Weaviate, or similar API design (FastAPI or equivalent), streaming, WebSockets Strong testing discipline — including evals for non-deterministic systems CI/CD, Docker, observability for LLM systems Nice to have ExperienceWhy it matters to usAnthropic / Google / OpenAI certificationsSignals you take this stack seriouslyLangGraph in productionIt's part of our stackSelf-hosted model ops (Ollama, vLLM, TGI)Core to our data-sovereignty storyPrivacy-centric or regulated-industry backendsAligns with Swisper's missionOpen-source contributions to agent / LLM toolingDemonstrates craft Engagement