AI Engineer – Agentic Systems

About the Position

We are looking for an AI Engineer to join our AI technology company, developing an intelligent inventory planning platform. The platform combines forecasting, optimization, and execution in one product, designed for customers across wholesale, retail, manufacturing, and food service. The core of our platform is a proprietary time series foundation model trained on millions of demand series and billions of data points across industries. We are expanding our engineering team and are looking for an AI Engineer to build the next generation of agentic capabilities around our forecasting and optimization technology.

Responsibilities

• Build the AI agent that customers will interact with every day.
• Connect language models with existing forecasting and optimization capabilities.
• Build the software that makes the agent useful, reliable, and safe.
• Define how the agent uses context, which models it should use, when it should take action, and when it should ask for confirmation.
• Feedback from real customer tasks will directly influence what you build next.

What You’ll Work On

Agent Harness

• Build the runtime around the model, including planning and tool-use loops.
• Manage session state, context, and memory.
• Implement retries, recovery, and resumable multi-step tasks.
• Provide customers with clear progress, results, and opportunities to intervene.

MCP and Tools

• Build Model Context Protocol (MCP) servers and integrations.
• Expose customer data, forecasts, and planning actions to AI agents.
• Design clear tool schemas with validated inputs and outputs.
• Implement access controls and confirmation flows for changes to customer data.

Context and Grounding

• Provide the agent with relevant customer data, business rules, and conversation history.
• Use retrieval and context management to keep responses traceable and useful as tasks become more complex.
• Keep calculations and business-critical logic in tested tools rather than relying on the model.

Evaluations

• Turn real customer tasks and failure cases into repeatable evaluation tests.
• Measure task completion, tool usage, and correctness.
• Combine automated checks, human review, and model-based graders.
• Identify regressions before shipping changes to prompts, tools, or models.

Model Selection

• Evaluate frontier models and open-weight alternatives against real workloads.
• Measure reliability, latency, and cost per completed task.
• Evaluate the impact of retries and hosting overhead.
• Make informed decisions about model routing, caching, and self-hosting.

Production Reliability

• Trace agent runs and investigate failures across models, tools, and data.
• Test permission boundaries, prompt injection, and unintended repeated actions.
• Own production rollout, monitoring, and continuous improvements.
• Work closely with product and customer-facing engineering teams.

Tech Stack

• Python
• FastAPI
• PostgreSQL
• Polars
• LangGraph
• LLM APIs
• MCP (Model Context Protocol)
• Claude Code

Requirements

• Strong Python and backend engineering skills.
• Experience building tested and maintainable backend services.
• Good understanding of APIs, asynchronous execution, and persistent state.
• Hands-on experience building and shipping LLM-powered applications.
• Experience with LLM tool use and structured outputs.
• Ability to improve AI-powered applications based on real user behavior and feedback.
• Understanding of agentic architectures, tool calling, and multi-step AI workflows.
• Strong problem-solving and ownership mindset.
• Comfortable working closely with founders and taking ownership of technical decisions.
• Professional English or German; either language is sufficient.

Nice to Have

• Experience with LangGraph or similar agent orchestration frameworks.
• Experience building MCP servers or MCP-based integrations.
• Experience with evaluation frameworks for LLM/agentic systems.
• Experience with RAG, retrieval, and context management.
• Experience with AI observability and tracing.
• Experience with prompt injection and AI application security.
• Experience evaluating and routing different LLMs.
• Experience with production AI systems and monitoring.

How We Work

We work in short development loops: identify a customer task, build a solution, evaluate it, and ship. You will discuss technical decisions directly with the founding team, share failures early, and help establish engineering and evaluation practices as the AI team grows.

We Offer

• Fully remote position.
• Opportunity to work on an AI product combining LLMs, forecasting, and optimization.
• High level of ownership and direct impact on the product.
• Close collaboration with the founding team.
• Opportunity to shape the architecture and engineering practices of an early-stage AI product.
• Work with modern AI technologies, including LLMs, agentic systems, and MCP.

About the Company

Our client is an AI technology company developing an intelligent inventory planning platform that combines forecasting, optimization, and execution in one product. The platform is designed for customers across wholesale, retail, manufacturing, and food service. At its core is a proprietary time series foundation model trained on millions of demand series and billions of data points across industries. The company is expanding its engineering team and is looking for an AI Engineer to build the next generation of agentic capabilities around its forecasting and optimization technology.