Senior Backend-Developer – Agents (100% Remote / German-Speaking)

PowerProzesse; United States; Remote; USD 22,000 - 45,000; Full time

Benefits offered by this job

100% remote work

Allianz Gold Health Insurance with top

Flexible hours

AI-tokens

Ownership and fast decisions

No unnecessary meetings

Job summary

Hausverwalter.ai is seeking a Senior Developer for its fully remote team. You’ll design and optimize autonomous AI agents that assist property managers, train templates, and improve how agents interact with tickets and knowledge bases.

You’ll work on the Co-Pilot, live agent tooling, and LLM infrastructure, selecting models, caching, and optimizing context. The role requires real-world AI production experience and strong German/English communication.

Qualifications

• German language proficiency required.
• 2+ years building production agents for real users.
• Experience shipping agentic features and tooling.
• Comfort talking to customers and turning feedback into product decisions.

Responsibilities

• Design and improve autonomous Inbox-Agents and templates.
• Define system prompts, skills, and prompts for agent behavior.
• Work on knowledge base integration and agent learning loops.
• Improve RAG search, tooling, and model selection for efficiency.
• Collaborate with customers to refine agent escalation rules.

Skills

German

English

Senior dev

Production agents

Tools

Convex

Next.js

Vercel AI SDK

React

TypeScript

Job description

Fluent German is required for this role. Our customers are property management companies in Germany.

We're looking for Senior Developers for our fully remote team to work on our agents. You'll be responsible for how well our agents work for customers: you talk to the property managers who use them, see where they go wrong and make them better.

About Us

Hausverwalter.ai is an all-in-one platform for property management companies with features like an Inbox, CRM, Documents, Telephony, Video Calls, Team Chat and AI-Agents. We want to automate as much of a property manager's daily work as we can.

About the role

Our platform has been an agentic platform since day one. Before we even planned the UI, we planned the way agents would interact with the software. Our goal is to automate as much of the work in property management as possible.

As of now, almost everything the user can do in the UI, we have an agent tool for it and the Co-Pilot (daily assistant of the user) can do everything for them, even toggling dark mode on and off.

But the Co-Pilot is just the beginning. The real challenge is the autonomous agents. More on that below.

What you’ll be working on

• Inbox-Agents: Autonomous Agents triggered whenever something new happens inside a ticket. The customer can train them on their own OR rather use one of the templates you will build for them. These agents already have the capabilities to fully own entire complex scenarios and are one of the main reasons customers are interested in our software. But when should the agent escalates to a human rather than finishing on its own? How can we make sure it sticks to the process? Those are your questions, and you answer them together with the customers who use these agents every day.

• Harness Engineering: The agent has 100+ tools. But it has to read a request the way a property manager would, not the way the words sound elsewhere. What goes into the system prompt, what into skills?

• Knowledge base: Our platform includes its own knowledge base, similar to Notion. It’s fully custom built and supposed to act as the companys collective brain. But should the AI also “learn” by itself? And if yes, when? And how can we improve the way our AI searches the knowledge base — through better tools, system prompts, or skills? When a customer asks us something in support, the answer should end up in the knowledge base so the agent can handle it next time.

• Voice-agents: The same agents, but on the phone — built on LiveKit, handling calls the way our text agents handle tickets.

• LLM Infrastructure. Pick the right model for the job, cache and stream to cut latency, retry when a provider fails, compact context, optimize RAG search, track usage across providers, reduce costs and run evals to see what's actually better.

Our minimalistic tech stack

Everything is TypeScript. From our database schema to the frontend.

Backend: Convex (Database & serverless functions)

Frontend: Next.js, React, Tailwind, shadcn/ui

Technologies you will encounter: Vercel AI SDK, Convex Agent Component

Salary

• $2000 - $4000 per month before taxes with room for growth and promotion in the months & years ahead

• Permanent employment contract in your country of residence through our partner Deel (EOR)

Benefits

• 100% remote work

• Allianz Gold Health Insurance with great coverage

• Flexible hours

• As many AI-tokens as you need

• We’re self-funded & profitable (and not burning VC money to survive)

• Small team, real ownership — what you build actually goes live

• Quick decisions, no pointless meetings and no layers of approval to get through

Our special Health Program

• Private room in hospital, worldwide coverage

• Free choice of doctors & specialists, no health check, pre-existing conditions don't matter

• Partner and kids are covered as well

• Surgery, cancer care, MRI/CT/PET scans and diagnostics fully covered

• Dental treatment & surgery covered at 90%

• $675/year for physio, osteopathy, chiropractic and acupuncture

• $100 per night in‑patient cash benefit, up to 20 nights

• Prescription drugs, vaccinations, glasses/contacts, home nursing and ambulance covered

Our values

• Get things done, every day matters

• Transparency and honesty, especially when it’s hard

• High standards for our own work and each other's

For a closer look at what we're building, check out our YouTube channel — heads up, it's in German.

You are:

• fluent in German and English.

• a seasoned developer with more than 2 years of experience building production agents

• AI-native in production. You've shipped agentic features real users rely on, and you have opinions on retrieval, evals, tool design, and where today's frontier models break down.

• a builder first. You'd rather ship a rough v1 today than spend three weeks planning the perfect one.

• up to date with new AI models & the research.

• writing almost all of your code with AI.

• used to working independently and taking full responsibility for your area, also after it goes live.

• comfortable talking to customers and turning what they tell you into product decisions.

• quick to learn new things and consider yourself a generalist, not a specialist.

Your techstack:

is not important to us, since AI will be writing 100% of your code.