Senior Software Engineer - Machine Learning

What We're Building

At caffeine.ai, AI doesn't assist the development process — it is the development process. We're building the infrastructure that makes self-writing software stable, fast, and production-ready. That means multi-agent systems orchestrating code generation in parallel, real-time pipelines streaming results to users as they happen, and databases holding the state of applications that were never manually written.

We're not building prompt wrappers. We're building the industrial-grade plumbing underneath them — and this role sits at the most critical seam: the layer that connects the agentic core to everything the user actually sees and touches.

What You'll Actually Do

• Own the SSE infrastructure that streams agentic job state in real time from backend to client — designing for latency, reliability, and graceful failure at every step
• Build and maintain the job orchestration layer that coordinates multi-agent workflows from dispatch through completion, including retries, state recovery, and context continuity
• Design database schemas and persistence strategies for long-running agentic work — jobs, artifacts, agent memory, and the user's evolving application state
• Build the bridge layer between raw agent output and the structured data models the frontend and other services dependon
• Instrument the full pipeline — latency, throughput, failure surfaces — and stay close to production behaviour across every release
• Collaborate directly with the AI orchestration team to evolve the contracts between agent logic and the infrastructure running beneath it

You Might Be a Great Fit If...

• You've designed real-time streaming systems in production — SSE, event-driven architectures, or similar — and you know where they fail under load
• You think about databases not just as storage but as a design surface — schema decisions, consistency guarantees, and state lifecycle are things you get opinionated about out
• You've worked with agentic or LLM pipelines in a backend context and understand the operational challenges of long-running, non-deterministic workloads
• You care about the user-facing effect of your infrastructure choices — latency, dropped events, stale state are product problems as much as engineering ones
• Ambiguity doesn't stall you — you scope the surface, make a call, and ship something you can measure
• You're energised by small teams where your work reaches real users within days, not quarters

Practicalities

• Location: Zürich, Switzerland — hybrid, 4+ days/week in the office
• Stack: TypeScript (Node.js/NestJS/Fastify), SQL/NoSQL, SSE, multi-agent architectures
• Team: Small, senior, high-trust — Core A.I.