AI Engineer – Agentic Systems (Voice) (m/f/d)

Your Role

• We are growing our
• Agents team
• and looking for an experienced
• AI Engineer (m/f/d)
• to help build the next generation of
• agentic AI systems
• powered by

Large Language Models (LLMs)

. Today the team is small and tight-knit, and every hire meaningfully shapes what we build and how we build it.

You will work across our full agentic surface: chat, meeting preparation, recommendations, and

voice

. This is a hands-on role at the intersection of reasoning, tool-use, and memory, designing intelligent agents that operate in complex environments, interact with APIs and structured data, and drive real-world outcomes. It goes well beyond prompt engineering: you will build robust decision pipelines and ship

product-ready agents

across both

text and voice modalities

.

Proven experience building

voice agents

is an important requirement for this role. Voice is one of the most exciting frontiers of our product, and we want someone who has done it to help the whole team excel there. That said, this is not a voice-only role: you will contribute across the Intelligence Layer, and your voice background will be a force multiplier for everything we build together.

Your Mission

Design & Optimize Workflows

• Design, implement, and optimize
• agentic workflows
• using LLMs across chat, meeting prep, and recommendations.

Build Tool-Using Agents

Create agents that interact with external APIs, databases, and environments.

Advance our Voice Agents

Push our

voice-enabled agents

forward, including pipeline design for real time conversation, transcription (ASR), speech synthesis (TTS), turn-taking, and dialogue management, with a focus on low latency and reliability.

Integrate Memory & State

Implement

retrieval, long-term memory

, and

• state tracking
• into agent architectures.

Deploy in Production

• Collaborate with product teams to
• deploy agents in production
• and keep them healthy.

Evaluate & Improve

• Design and run
• experiments
• to evaluate agent performance, correctness, latency, and hallucination rates.

Translate Product Needs

• Turn high-level product requirements into scalable, testable agentic features.
• Your profile: What you need to succeed

Voice AI (important requirement)

• Proven, hands-on experience building
• voice agents
• in production and working with

ASR

,

TTS

, or real-time/streaming voice pipelines. Voice is a core focus of this hire, and we expect you to bring real depth here.

LLM Expertise

Strong experience building with

LLMs

(e.g., OpenAI, Claude, Mistral, open-source models).

Agentic Frameworks

Familiarity with

• agentic frameworks
• (e.g. LangChain, LangGraph).

Programming Proficiency

Proficient in

Python

and modern ML tooling (e.g., Hugging Face, PyTorch, Ray, etc.).

Tool Integration

Experience with

• tool integration
• , calling APIs, managing memory/state, or SQL query generation.

Simulation-Based Testing

Experience with

• simulation-based testing and evaluation
• of conversational and voice agents (e.g.

Coval

,

Bluejay

), used to measure agent performance, correctness, and regressions at scale.

Startup Mindset

• Comfortable in an iterative, fast-paced product environment, and energized by helping a small team level up.
• Why Acto?

Development Opportunities

: Steep development opportunities without entrenched hierarchies.

Culture

We are an ambitious team that wants to achieve a lot with Acto, but we don't take ourselves too seriously and like to have fun together, be it working together or at regular team events.

Flexibility

Enjoy a flexible work schedule and a remote set up with the option to work from our Berlin Office.

Health & work-life balance

Enjoy 30 days of vacation and an attractive Wellhub membership.

Ownership

Whether you are an intern or a full-time acto-naut, Acto instills a deep sense of ownership, where every team member is entrusted with meaningful responsibilities and the autonomy to make impactful decisions.

Equipment

You will receive a state-of-the-art hardware setup with a choice between Mac and Windows.