Qriosity Labs home

Bring us the question. We'll build the answer.

We build AI agents and knowledge systems that your risk, security and engineering teams can govern, audit and own.

Start a project See our work

most projects start with a 20-minute call

What we build

  • The bridge of a ship, with rows of monitors and control panels

    Governed AI agents

    Agents that investigate, decide and act through controlled tools. Consequential actions wait for a person.

    See the agent projects
  • Rolling archive shelves in a records store

    Auditable knowledge systems

    Retrieval and knowledge graphs where each answer traces to a source or a rule.

    See the knowledge projects
  • A crate of tomatoes at different stages of ripeness

    Applied machine learning

    Vision, fraud and prediction models, reported on held-out data with stated limitations.

    See the ML projects

See all services

How a project runs

  1. Question

    We dig into what you actually need, not just what you asked for.

  2. Explore

    Quick prototypes and options, so you can see ideas before committing to them.

  3. Build

    Weekly check-ins and working demos. No disappearing for months.

  4. Ship

    Launch, measure, and improve. We stick around after go-live.

Lab notes

Recent experiments that made it out of the lab.

  1. Diagram: a planner hands work to an investigation loop. A critic sends weak findings back for remediation, and a deterministic verifier checks the result before it becomes a report in which each claim cites its evidence. ERP data, public filings and internal documents feed the investigation through read-only access. code runs the process remediate Plan Investigate Critique Verify ERP Filings Docs read-only, guarded access Report each claim cites evidence

    Governed agents

    Autonomous financial investigation platform

    Explains why profitability moved, with an evidence trail an auditor can follow.

    Read the note
  2. Diagram: six agents share one pipeline. They gather, assess, validate and decide, then a person approves before the agent executes and verifies the change by reading it back. Every tool call is written to an audit log. six agents, one pipeline Gather Assess Validate Decide Approve Execute Verify Audit log a person approves read-back every tool call

    Governed agents

    Operations automation agents

    Six agents that investigate a business event, decide, act through controlled tools and escalate to a person when required.

    Read the note
  3. Diagram: a piece of produce with two defects outlined and measured. Background removal and four segmentation methods find each defect, and an ontology reasoner returns a diagnosis together with its reasoning. size · colour · coverage Remove background Segment (4 methods) Ontology reasoner Diagnosis + reason

    Applied machine learning

    Explainable visual inspection

    Finds, measures and explains defects on produce, and shows the reasoning behind each diagnosis.

    Read the note

See all 15 lab notes

The curious people behind the lab

We're a small team of designers and developers who like hard questions. We ask a lot of them early, then build in small steps so you can see real progress every week.

Meet the team

What are you building?

Tell us what you're working on and we'll get back to you within two working days.

Start a project