Applied research consultancy

Applied research that brings
digital systems to light.

and the tools, when a project needs them.
Light Light Labs

We take on research questions about digital technology: how systems behave, who they disadvantage, and whether they do what they claim. Most engagements produce a study and a report. When findings need to be put into practice, we build the tool in house.

Start a project → What we do
Applied research✳ Study design✳ Bias & fairness audits✳ Data analytics✳ In-house development
01The practice

Every project begins as a research question.

Some end there, with a report and a dataset. Others end with working software, because a finding only matters once it is applied. We build that software ourselves rather than handing over a specification and hoping it survives.

We work with public agencies, non-profit organisations, universities and research teams.

reports include method, data and limitations.
02What we do
01

Applied research

Study design, data collection, analysis and reporting.

most projects stop here.

We work on questions about digital systems: algorithmic bias and fairness audits, evaluation of AI tools, and program or policy questions where the evidence matters more than the opinion.

Every report states the method, the data and what the findings do not establish. A research engagement is complete on its own. Nothing here obliges you to commission software afterwards.

You receive a written report, the dataset and the analysis code behind it, and a briefing for the people who have to act on it.

  • Literature review
  • Study design
  • User research
  • Bias & fairness audits
  • Quantitative analysis
  • Evaluation
  • Reporting
Bring us a question →
02

In-house development

When a project needs a tool built.

only when it's needed.

A data pipeline, an audit instrument, an internal application, an automation. When an engagement needs one, we build it here, so the people who ran the study are the ones specifying what gets made.

We also take on development work on its own, where the research is already settled and the job is to build the thing.

  • AI automation
  • Data pipelines
  • Audit instruments
  • Applications
  • iOS · Android · Web
Tell us what you need built →
03Selected work
01

Statelessness Knowledge Platform

Research, data and resources on statelessness, in one place.

funded by UNHCR.

A knowledge platform gathering research, data and resources on statelessness and making them usable by the people who work on it.

Funded by UNHCR
  • Knowledge platform
  • Research
  • Applied policy
02

Unbordered Knowledge

Responsible AI training.

training programme.

A training programme on using AI responsibly: what these systems actually do, how they fail, and how to judge their output instead of taking it on trust.

Sessions cover practical use, the limits of current models, bias and fairness, privacy and data handling, and how to set internal rules people will follow.

  • Responsible AI
  • Training
  • Bias & fairness
  • Workshops
04Our own products
01

Sprout World

A learning app for children aged 3 to 6.

coming to iPad.
Sprout World

Seven places to explore, a companion character, and a garden that grows as a child completes activities. No points, no timers and no in-app purchases. It collects nothing — no accounts, no ads, no trackers. All data stays on the device.

  • Ages 3–6
  • iPad
  • Works offline
  • No data collection
Visit Sprout World →
02

Scholar OS

A dashboard for graduate research.

in development.
Scholar OS

A graduate degree is run across a dozen tabs: papers to read, drafts in progress, calls for papers, conference deadlines, and whatever your supervisor said three weeks ago. Scholar OS puts them on one dashboard.

Agents work alongside it — watching for new literature in your field, tracking open calls and deadlines, and available to think out loud with when a chapter stops making sense.

  • Literature tracking
  • Calls & deadlines
  • Research agents
  • Drafts
Request early access →

Scholar OS, running

click around →
Scholar OS
Fields Algorithmic fairness Human–AI interaction
This week

A working slice of the dashboard, with sample data. Mark papers as read, track a call, run the literature scan, or ask the brainstorm agent something.

05Writing

What we are working through at the moment. We publish method and limitations alongside conclusions, so the reasoning can be argued with.

01

What fairness metrics miss when the ground truth is itself biased

in progress

Most fairness measures compare outcomes against recorded labels. When those labels were produced by the same process being audited, parity can be satisfied while the harm continues.

02

Auditing a model you cannot see inside

in progress

What can and cannot be established about a system when you have access only to its inputs and outputs, and how to say so precisely in a report.

03

How people actually use AI to answer questions that matter to them

in progress

Field observations on what people ask, what they believe, and where a confident wrong answer does the most damage.

06Expertise

Light Light Labs works across doctoral-level research on bias in AI systems, user experience research, data analysis and software engineering.

That range is what lets a single engagement run from a research question through to a working tool, without the work changing hands partway.

07Contact

Get in touch.

hello@lightlightlabs.com