Construction doesn't buy "AI". It buys answers it can act on — and an answer can look completely plausible while being wrong in the way that costs money on a real project. The Construction Evaluation Engineer role exists to catch exactly that: to define what great looks like for our agents, find where they fail, and turn that judgment into something our systems can measure.
It starts hands-on and close to the data, and gets more technical as you grow — from evaluating outputs to owning eval pipelines, graders and datasets. You'll work alongside our forward deployed engineers, design partners and construction experts. Evaluation here happens close to the customer, not in a disconnected labeling operation.
**What you'll do**
* Run our agents against real construction workflows and judge the outputs on correctness, completeness and practical usefulness — not on whether they sound convincing
* Turn real workflows, edge cases and production failures into repeatable eval cases: reference outputs, scoring rubrics, failure taxonomies and expert-reviewed datasets
* Grow into the systems around the work — eval harnesses, automated graders, regression tests, model comparisons
* Work out what genuinely needs human judgment, then engineer the rest away
* Capture what we learn from customers, experts and agent failures so every deployment improves the underlying system
**You might be a fit if**
* You have real exposure to construction, engineering, project planning, estimating or surveying
* You're strongly quantitative and analytical, with a serious interest in modern AI systems
* You're extremely detail-oriented — you can work through hundreds of cases without letting the quality bar drift
* You're curious about _why_ systems fail, not just whether they fail
* You have an instinct for automation: understand the manual workflow first, then build the system around it
* You have high ownership — you find the issue, understand it, and close the loop
* You speak Swedish and English
You don't need to already be an experienced AI engineer. You need the raw ingredients to become one.
**What success looks like**
* We know where our agents are weak before a customer finds out
* "Good output" stops being a matter of opinion and becomes something we measure
* Failures turn into eval cases instead of recurring surprises
* You grow into a larger technical seat — Evaluation Engineering, Applied AI, Forward Deployed Engineering or AI/Data Engineering. This isn't a permanent part-time labeling role; the long-term role gets built around you
**The practical details**
Part-time to start, roughly 15–20 hours per week, flexible on how they're distributed. Hybrid with meaningful time at the office at Linnégatan 89B, Stockholm. Paid hourly, growing with the company in responsibility, compensation, ownership and equity.
**Process**
Intro call, 15–20 minutes. A short home assignment that reflects the actual work. Then come to the office and meet the team. Fast process, quick feedback at every stage.
Keen on doing your life's work — bringing one of the world's largest and most impactful industries, the $15 trillion construction industry, into the AI era? Send us an email at careers@ brickanta.com or apply directly at .
**About Brickanta**
Brickanta ([Brickanta.com](http://Brickanta.com)) is the AI-native agentic platform for construction bid analysis, cost estimation and procurement. Brickanta's industry know-how AI raises margins and saves time by leveraging historical estimates, tender documents, and project data — so builders can focus on building. The founding team has been building with AI since 2018 and brings deep experience applying AI in construction and industrial sectors. Brickanta has raised over $8M in external funding from leading investors, including Northzone, Anton Osika and Mario Götze.
Visa sponsorship is offered for this role.
