Engineering Intern

Atira GmbH; München; On-site; EUR 13,000 - 20,000; Full time

Benefits offered by this job

Competitive salary

Mentorship

Additional benefits

Job summary

Atira GmbH in Munich is seeking an Engineering Intern to work on frontier AI technology, including agent runtimes, harnesses, and evaluation pipelines across the sales-to-ERP lifecycle.

You will build frameworks and tooling, improve the context layer, and collaborate with FDEs and GTM teams to ship features that impact customers and industrial partners worldwide. This is a hands-on three-month internship with real ownership.

Qualifications

• Ownership, depth, and accountability for解决 complex systems.
• Solid Python/TypeScript programming skills and AI exposure.
• Experience with agents, evals, runtimes and tooling is a plus.

Responsibilities

• Work on the agent runtime and harnesses to scale decision making.
• Improve the ontology and evaluation pipelines for reliability.
• Scale multi-modal industrial data handling across ERP-to-CRM.
• Collaborate with FDEs and GTM to ship customer-facing tooling.
• Provide clear ownership of end-to-end components.

Skills

Python

TypeScript

AI agents

Runtimes

Education

Computer Science

Tools

Sandboxing

Job description

About Atira

Atira builds the commercial brain for industrials. Our agents 10x the sales engineers of today by offering the complete toolbox of tasks needed in complex environments. They handle everything from extracting and classifying requirements from incoming RFQs, orchestrating communication and information retrieval across multiple stakeholders and systems, while generating outcomes like documentation, production system inputs, or configuration recommendations. CRM. Atira. ERP.

We are one of the fastest growing teams in Munich and are looking for you to join us!

Why engineers are the foundation of everything Atira ships

As an Engineering Intern at Atira, you will work on frontier AI technology, such as (recursive) agent harnesses, runtimes, evals, inference and self-verification in complex engineering domains. You build the frameworks, primitives and scaffolding that let agents reliably handle complex industrial processes end to end, from the first inbound RFQ to the write-back into ERP.
Misreading or wrongly interpreting a spec can cost a manufacturer six figures, so the challenges are real and the margin for error is thin.
Where you trained matters less than what you've built and how clearly you can describe what broke. Atira is at a stage where you must own problems end to end, because what you ship today is what our customers use tomorrow.

What you will do

• Work on the agent runtime and harnesses. You improve the core components that let Atira's agents reason in complex engineering environments across the entire sales-engineering lifecycle, orchestrate themselves recursively at scale through dynamic programs, and handle arbitrarily large amounts of data. This includes the runtime/sandboxing, the harness, and the outer orchestration layer that ties it together

• Build out our AI moat. You are in charge of improving quality & evals, improving the ontology layer that unlocks compounding process knowledge and fusing deterministic guardrails with agentic agency to make our agents actionable at enterprise level complexity

• Unlock industrial scale. Industrials live in multi-modal, data-heavy environments: 1000-page inbounds, pricing lists, CAD files, ERP tables, legacy applications. You scale the AI systems behind them to millions of concurrent LLM requests in isolated sandboxes, with no silent failures, cost blowouts or latency explosions.

• Advance our internal tooling. FDEs, platform engineers and the GTM team depend on tooling you build to improve Atira's internal context layer, ambient agents, insight and observability engines

• Work with FDEs when customer problems hit the platform. When an FDE hits a problem they should be able to pull you in. At a team of 20, the boundary between AI, platform and customer is thin, and you'll cross it regularly

What we are looking for

Ownership, depth, and the instinct to take problems personally, go deep until you understand the system, and own the outcome including the parts that broke.

Your profile

• Agent fluency sits on top of real engineering fundamentals. Solid skills in Python, TypeScript, or a comparable language. We value someone that can still code without AI. Ideally paired with hands-on experience with agents, evals, harnesses, runtimes, sandboxing and tools

• You've gone further than most with AI-assisted development. Whether it's Claude Code, Cursor, Codex, or your own setup, you've invested seriously in building the skills, context, and workflows to get maximum leverage from coding agents. This isn't a nice-to-have. We think engineers who have mastered this are fundamentally more productive, and we hire accordingly

• You reason clearly about technical tradeoffs. You don’t need to have run systems at scale yet, but you can think through how a design would hold up at thousands of requests per second and explain the pros and cons of different approaches. Every design has tradeoffs: no free lunch

• You’ve demonstrated ownership and initiative. A side project you took further than anyone asked, a serious open-source contribution, a research project or a startup. We want evidence that you build without being told to and can explain what broke and why

• You have a strong technical foundation. Enrolled or recently graduated in Computer Science, Information Systems, AI, or a related technical field from a strong technical university (cracked dropouts can qualify too though)

• Full working proficiency in English. German is not required

• You're ready for a real internship, not a side gig. You can commit to at least three months full-time, on-site in Munich

What you'll get

• Real ownership: you'll be one of the early team members and shape both product and architecture as well as work closely with customers

• Strong peers and mentorship: work with and learn from engineers who have shipped AI systems in industrial environments at scale, with the opportunity for mentorship along the way

• Impactful work: your code is deployed all across Europe’s and US industrial backbone, affecting how complex products and services are sold and configured worldwide

• Competitive salary and additional benefits

• Most importantly: Collaborate & thrive in a high-performing but caring culture. We will own the industrial sales function but will do so with modesty and integrity

What success looks like at the end of the internship

• You've shipped meaningful improvements to the agent runtime, our AI pipelines and related systems that are running in production across multiple customer deployments

• You own one well-defined piece of our AI stack end to end (an eval suite, an internal tool, part of the context layer or the SDK), and the team trusts your judgment on how it should evolve

• When something breaks in the part you own, you can trace it end to end, from agent behaviour back to the agent infrastructure layer, and fix it

• You’ve made an improvement to our internal AI tooling that the team now uses day to day

• The design decisions in your piece have held up under real load, and you can explain the tradeoffs you made and what you’d do differently

• If these metrics are met, we'd take you full-time without thinking twice