Senior Machine Learning Engineer

We are partnering with an early-stage AI startup building the infrastructure to turn human computer work into software. Their product captures how work is actually done — across screens, tools, and systems — and converts it into structured knowledge that autonomous agents can execute on.

They have raised $4m at pre-seed from a highly intentional investor base, including backers behind UiPath, Revolut, and Neo4j. The product is already live with paying customers, with early users seeing measurable ROI — including hundreds of hours of manual work removed each month.

What you will do:

Build retrieval and memory layers that allow agents to reason over large-scale organisational data;

Transform multimodal inputs (screens, keystrokes, system data) into structured knowledge graphs;

Work hands-on with vision-language models and LLMs in real production systems;

Define and own evaluation frameworks to measure agent quality, reliability, and improvement;

Run applied ML research and translate it into production-ready features;

Collaborate directly with the founders to shape the ML and overall technical direction.

What they are looking for:

Strong applied machine learning background, with experience shipping models into production;

Hands-on experience with VLMs, retrieval systems, or RAG architectures;

Comfort working with graph structures, ontologies, or knowledge representations;

Track record of running rigorous experiments and improving model performance;

Ability to balance research with execution — this is not a pure research role;

Comfort operating in a high-ownership, early-stage environment with minimal structure.

Tech stack

Python, TypeScript, React, FastAPI;

Postgres, Temporal, Electron;

AWS, Docker, Terraform;

VLMs, retrieval systems, knowledge graphs, ontology modelling.

Package

Up to EUR 160,000;

Equity at the top end of the range for senior profiles;

Munich-based, with full in-person collaboration (relocation supported).

Why join

You will be joining at pre-product-market-fit, working directly with the founders to define how AI agents actually operate in real-world environments. This is a zero-to-one role where you are architecting the core ML systems — not maintaining them.

From a career perspective, this role offers the opportunity to define the ML direction behind the company’s world model — arguably the most valuable and technically rich part of the stack. You will own problems that combine multimodal data capture, ontology-aware retrieval, and continuously improving agents, end-to-end.

The technical challenge is non-trivial. This is not an off-the-shelf RAG system — it sits at the intersection of VLMs, graph-based reasoning, and agent evaluation, with open questions still to solve around fidelity, scale, and reliability.

You will be working directly with the founding team, including a highly technical CTO, with investors behind Neo4j reinforcing the long-term graph-native direction. The team is small, high-calibre, and operating with urgency — the expectation is high intensity and deep ownership, but the upside is building something category-defining from first principles.