AI Engineer – Agentic Systems for B2B Outbound

Startupticker; Zürich; On-site; CHF 120,000 - 180,000; Full time

Job summary

Arcton, a Zürich-based AI native agency, seeks an AI Engineer to build agentic outbound systems for B2B. You’ll design architecture for tool calling, function routing, and multi-step plans that operate across several specialized agents.

You will work with Python, LangChain, LangGraph, and OpenAI function calling, ensuring reliability for live client deployments and handling failure modes in complex pipelines.

Qualifications

• Proficient in Python with production-grade software discipline.
• Experience designing and debugging agent loops and multi-step workflows.
• Knowledge of tool schemas, function routing, and state management.

Responsibilities

• Design and implement agent loops and task planning.
• Structure tool and function schemas for reliable LLM calls to external systems.
• Ensure reliability, observability, and fail-safe behavior in live client deployments.

Skills

Python

LangChain

LangGraph

OpenAI function calling

Agentic systems

System architecture

Tool orchestration

Tools

QuickMail

GetSales.io

Apollo

Prospeo

Job description

AI Engineer – Agentic Systems for B2B Outbound

Arcton is a Zürich based AI native agency building autonomous outbound systems for B2B companies. Our core product is an AI sales agent that clients brief like an employee, no dashboards, no manual setup, and it runs prospecting, sequencing, and qualification end to end.

We're looking for someone who actually understands how agentic systems work in production, not just how to call an LLM API. You'll work directly on the architecture behind our AI agent: tool calling and function routing logic, multi step task planning, memory and state handling across long running workflows, and orchestration between multiple specialized agents (one handling lead research, another handling message drafting, another handling deliverability checks, for example).

Concretely you'll need to be comfortable with designing and debugging agent loops (not just prompting a single turn chatbot), structuring tool and function schemas so an LLM can reliably call external systems like CRMs, email infra such as QuickMail or GetSales.io, and enrichment APIs like Apollo or Prospeo, handling failure modes in agentic pipelines such as hallucinated tool calls, infinite loops, or bad state transitions, and basic RAG or retrieval setups for grounding agent outputs in real client data. Python is required. Experience with LangChain, LangGraph, Claude or OpenAI function calling, or building anything agentic from scratch (even a side project) is a strong plus.

This isn't a prompt engineering role. You'll be shipping things that run live for paying clients, so you need to think in terms of reliability, observability, and what happens when the agent gets it wrong.