Agentic Platform Engineer (gn) @ AI Venture, Berlin

This is an Atlantic portfolio venture. About the Venture AI models have moved faster than most companies’ ability to put them to work. The capabilities are there; knowing how to combine them effectively is still catching up. Frontier labs naturally put their own models first. Companies often follow suit, asking one general-purpose model to handle ten very different jobs. We believe there is far more to unlock in the models that already exist, even before the next generation arrives. That is the opportunity we are building around. We bring closed providers and open-weight models together behind a single API. Our engine breaks requests into their component parts, routes each to the right model within a cost budget, fine-tunes on customer data and checks the result against agreed quality standards. With no proprietary model to promote, we can choose the right combination for the task. Lower costs matter because of what they make possible: giving a small team the freedom to build something that would previously have required a much larger one. Our founder brings considerable drive, a hands-on approach and a strong sense of urgency to building the company. Backed by Atlantic at pre-seed, we are growing the engineering team to bring the product into production, with people already waiting to use it. About the Role As one of our first platform engineers, you will build the foundations that every product feature depends on: the gateway and router, usage and billing systems, customer key management and the European infrastructure underneath. Much of that platform is still taking shape. Working directly with the founder, CTO and founding AI engineer, you will help turn early architectural decisions into systems that remain reliable as the product grows. You will take capabilities from design through to production, thinking through their APIs, failure modes, deployment, observability and ongoing operation. In a team this small, building and running the system belong together. You will also work alongside our founding AI engineer on evals, embeddings and open-weight model serving, bringing platform engineering depth to the ML side of the product. Agentic engineering is central to how we work. You will operate in an agentic way by default, using coding agents as a core part of how you design, build, test and ship software. You'll develop the harnesses and workflows that allow you to delegate and parallelise work effectively, increasing what a small engineering team can deliver. This requires strong coding fundamentals and the judgement to review, challenge and take responsibility for everything that reaches production. There is a lot to build and an ambitious timeline ahead. We work with intensity, move quickly from decisions to execution and value people who take initiative and see things through. What You’ll Work On Your first area of ownership will depend on your strengths and our priorities. From there, you will contribute across the platform as it develops. • Gateway and routing: build a consistent API across model providers, with reliable streaming, fallbacks and safe handling of retries and provider errors. • Agentic engineering workflows: build and refine the harnesses, tooling and workflows that enable coding agents to work effectively across the development lifecycle, from implementation and testing to debugging and deployment. • Cost optimisation: develop caching, history compression and model selection, measuring savings while keeping estimated costs distinct from actual spend. • Usage and billing: build accurate accounting for model calls, usage reports, plan entitlements and invoice reconciliation. • Agent infrastructure and integrations: support tool calls, cost limits, human approvals and runs that can recover safely from interruptions, alongside connectors for services such as Gmail, Slack and calendars. • Tenancy and secrets: maintain workspace isolation, secure authentication and encrypted customer key management. • European infrastructure and model serving: build and operate our cloud infrastructure, including the GPU capacity used to serve open-weight models. • Production reliability: make systems observable, migrations reversible and releases safe, with tests that protect data integrity and logs that keep customer content and secrets private. The platform will evolve quickly. You will need to move comfortably between improving what exists and learning enough about a new problem to build the next piece well. Our Stack Our backend is built in Python, with a TypeScript frontend in a shared monorepo. • Backend: Python, FastAPI, Pydantic and LiteLLM. • Data and queues: Postgres with row-level security and pgvector, Redis and arq. • Frontend: Next.js, React and TypeScript. • Cloud and infrastructure: Scaleway, Terraform and Kubernetes for GPU workloads. • Auth and secrets: Zitadel, Secret Manager and KMS. • Observability and testing: OpenTelemetry, Langfuse, Sentry, pytest and testcontainers, alongside frontend testing tools. You should be comfortable with the core backend technologies and ready to learn the rest as the work requires. About You You enjoy the craft of engineering and the responsibility of seeing your work used in production. Faced with an unclear problem, you can identify what matters, make sensible trade-offs and carry a solution through to release without someone planning every step. You naturally look for ways to multiply your output through agentic workflows, rather than approaching every engineering task as something to execute manually. We are looking for: • Experience building and operating production systems: several years of backend engineering with real users, typically three to six, and meaningful ownership of what you shipped. • Strong Python fundamentals: confidence with async code, typing and testing. TypeScript experience is a plus. • Agentic engineering as a default way of working: you have hands-on experience using coding agents as a core part of your engineering workflow. You know how to break down problems, delegate and parallelise tasks, provide the right context and build effective agentic workflows. Crucially, you have the technical depth to review, debug and take ownership of the resulting code. • Platform engineering depth: practical experience with Postgres, Redis, queues, networking and Terraform. Kubernetes experience is useful, particularly for our GPU infrastructure. • Sound engineering judgement: clear APIs, tenant isolation, data integrity, partial failure and safe rollouts. You can explain both your decisions and what you chose to leave out. • A working understanding of ML systems: familiarity with LLMs, embeddings, evals, model serving and fine-tuning. Experience with LLM APIs, streaming and token costs is a strong plus. • A disciplined approach to measurement: you benchmark changes, question assumptions and use evidence to understand whether an improvement holds up. • Initiative and drive: you enjoy getting deep into a problem, finding a way forward and putting in the work to make it happen. • This is an experienced individual-contributor role, with ownership of the capabilities you build and close collaboration across the founding team. • Fluent English is required; German is a plus. Why Join At this stage, platform engineering decisions have a direct influence on the product and how quickly the team can build it. You will work close to the founder and CTO, with room to shape the systems you own and see the results of your work first-hand. The challenge spans AI infrastructure, distributed systems, security and cost efficiency. You will build across those boundaries, take real responsibility early and receive a meaningful equity package in the company you help develop. We are competing with teams that have more people, more funding and an earlier start. Our approach depends on technical depth, focused decisions and the drive to turn them into working software. We believe agentic engineering can fundamentally change what a small, technically strong team is capable of building, and we're looking for people who want to push that approach forward. For someone who enjoys a demanding build, close collaboration and visible impact, there is a lot to get involved in. This is a full-time role, based in our Berlin office five days a week, starting as soon as possible. Find more English Speaking Jobs in Germany on Arbeitnow