Staff Software Engineer - AI

tl;drAfter 14 years in construction tech, Lea & Jonas are building the AI foundation for the construction industry 6 of Europe's 20 largest construction companies already rely on Kuro, with leading construction enterprises representing over €40bn in combined annual construction volume using the platform every day Construction has some of the most complex, unstructured data you can throw at an LLM. At Kuro, you can tackle some of the hardest problems in applied AI every day Curious about the bigger engineering challenge? Explore the vision and open problems at kuro.build. About usLea and Jonas scaled two of Europe's leading construction tech companies - Capmo and alcemy - and saw firsthand the challenges holding the industry back. Now they're tackling the root cause: the 50% of construction work still lost to manual data handling. Kuro is building the AI foundation for construction. We're reimagining how construction professionals get work done and make decisions by leveraging AI to augment back-office workflows and make core data accessible. We are not pursuing marginal improvements, but a step change for construction. After one year on the market, leading construction enterprises with a combined annual construction volume of >€40 billion use Kuro every day to automate workflows and make faster, better decisions. Backed by an €10 million seed round, we already process more than 1 million pages every month and expect that volume to grow 10x over the next six months. We are 15 people based in Berlin. You will join a group of technical ex-founders and builders who are ambitious and kind. Learn more: Join Kuro - Berlin What You'll DoSet the technical direction Own the architecture decisions for the platform, from document ingestion to the chat UI Write and review RFCs. The technical roadmap for the next two to three quarters is yours. Make the build-or-buy decisions: which models we use, and which retrieval and orchestration stack we build on. Turn a vague customer request (“show me where the bid deviates from the BoQ”) into a scoped plan with the risks named. Raise the bar for the team Mentor engineers: Review their designs and their code. Define the engineering standards. Find what slows the team down, in tooling or in the CI/CD pipeline, and fix it. Interview and hire engineers. You have no direct reports. You change decisions in product and in other teams with a written argument. Build and own end-to-end (AI) features Design and implement new features to solve a specific user problem, such as document versioning or our agentic Chat. From concept to DB schema, to optimizing LLMs, to UI. Own the full development lifecycle of customer-facing features. Manage evaluations and scoring, and perform error analysis on new and shipped AI features. Build the platform that abstracts workflow building Contribute to the workflow harness, decision-trace standard, and eval tooling so a new use case ships in ~1 month Build reusable human review & approval surfaces where trust in AI output is earned Work on the construction ontology & project graph Help model and evolve canonical entities of our customers Contribute to entity resolution across sources: recognizing the same requirement, position, or price across BoQs, contracts, drawings, and bids Build features so answers are evidence-linked and auditable, and decisions flow back into the graph Contribute to scalable data pipelines & integrations Extend our multi-modal ingestion of unstructured documents - PDFs, scans, drawings, and 3D models - orchestrated and cost-efficient at scale Build read/write integrations with customer systems (document management, ERP/AVA), so workflows complete work end-to-end Projects you would be involved in Current & Past projects Project & document versioning guaranteeing data lineage in construction projects German tokenizer for construction-specific search Document pipeline ingesting 10,000s of PDFs per day Native BIM/DWG