Technical Co-Founder (AI Efficiency) @ AI Venture, Berlin

This is an Atlantic portfolio venture.

About the Venture

AI models have moved faster than most companies’ ability to put them to work. The capabilities are there; knowing how to combine them effectively is still catching up.

Frontier labs naturally put their own models first. Companies often follow suit, asking one general-purpose model to handle ten very different jobs. We believe there is far more to unlock in the models that already exist.

We bring closed providers and open-weight models together behind a single API. Our engine breaks requests into their component parts, routes each to the right model within a cost budget, fine-tunes on customer data and checks the results against agreed quality standards. With no proprietary model to promote, we can choose the best combination for each task.

Lower costs matter because of what they make possible: giving a small team the freedom to build something that would previously have required a much larger one.

Our founder was selected for Slush 20 Under 20, awarded by Anthropic, Slush, Index Ventures and Sifted. We closed our pre-seed round in two weeks and already have people waiting to use the product. We're now looking for a technical co-founder to help turn this opportunity into a company.

About the Role

As Technical Co-Founder, you will lead the development of our AI efficiency engine, working alongside the founder to turn a compelling technical opportunity into a product.

At the heart of the role is a deceptively simple question: how can we make AI models deliver the same or better results at a fraction of the cost?

You will explore approaches to model routing, fine-tuning, caching and inference optimisation, combining scientific curiosity with the ability to turn promising ideas into measurable improvements.

This is a role for someone who enjoys difficult technical problems, rigorous experimentation and challenging existing approaches. You won't be expected to build every part of the platform yourself. Working alongside our founding AI engineer and backend platform engineer, you will set the technical direction and focus your expertise where it creates the greatest impact.

Agentic engineering is central to how we work. We're looking for someone who already operates this way, using coding agents and developing the workflows and harnesses that allow a small team to move exceptionally quickly without compromising technical quality.

Above all, this is a co-founder role. We're looking for someone who wants to take genuine ownership, work with relentless intensity and help build a company from its earliest stages.

What You'll Work On

• Drive AI cost efficiency: develop approaches to reducing inference costs through model selection, routing, caching, fine-tuning, quantization and other optimisation techniques.

• Experiment and benchmark: design rigorous evaluations to understand the trade-offs between cost, quality and latency, turning technical insights into measurable improvements.

• Shape the efficiency engine: determine how closed and open-weight models can be combined and optimised to deliver the best outcomes for different tasks.

• Set technical direction: identify promising opportunities, make key architectural decisions and guide the founding engineers in bringing them to production.

• Push agentic engineering forward: use coding agents as a core part of your work and develop the harnesses and workflows that allow a small team to operate with exceptional leverage.

• Build the company: work closely with the founder on technical strategy, attract exceptional talent and help shape the product and business.

About You

You bring deep technical curiosity, scientific rigour and an entrepreneurial mindset. You enjoy getting into the details, testing hypotheses and finding approaches others might overlook.

We're particularly interested in people who bring:

• Deep expertise in AI efficiency: meaningful experience optimising the cost or performance of LLMs or other AI systems, with concrete results to show for it. Experience in inference optimisation, routing, fine-tuning or related techniques is particularly relevant.

• Strong technical foundations: a solid understanding of modern AI models, rigorous experimentation and the ability to translate research into practical improvements.

• An agentic approach to engineering: you actively use coding agents, know how to orchestrate them effectively and are excited by what small, highly leveraged teams can achieve.

• A co-founder mindset: you're resourceful, ambitious and comfortable with uncertainty. You take initiative, move quickly and want to build something from the ground up.

• Technical leadership potential: you have the judgement to make difficult technical decisions and guide others. Previous CTO or engineering management experience is not required.

We're more interested in exceptional depth in AI efficiency and the ambition to build a company than a conventional CTO background.

Fluent English is required; German is a plus.

Nice to Have

• Experience with open-weight models, GPU inference or model serving.

• A strong research or open-source track record in AI or machine learning.

• Experience taking AI systems into production or working in an early-stage startup.
  

Why Join

With the pre-seed round closed and early momentum behind us, you'll join at a point where there is considerable room to shape what comes next. You'll work alongside the founder, set the technical direction and receive meaningful co-founder equity in the company you help create.

We're competing with teams that have more people, more funding and an earlier start. Our conviction is that exceptional technical depth, focused decisions and an agentic approach to engineering can give a small team a meaningful advantage.

For someone who wants to apply their expertise in AI efficiency to an ambitious technical challenge, and take the leap from specialist to co-founder, this is an opportunity to build something of their own.

This is a full-time commitment, based in our Berlin office five days a week, starting as soon as possible.