CountingLab

Rapid AI Prototype development

Pioneering AIto empowerthe future

We turn advanced research methods into commercial AI: computer vision, forecasting and time-series prediction.

See our work Start a project
LiveGB grid carbon intensitygCO₂/kWh
5010015020025030035018:00Mon06:0012:0018:00Tue06:0012:00now · forecast →← actual
Now (forecast)264 very high
Last reading256 at 15:00
Forecast error, 24h±6.3 MAE
Cleanest next 24h192 at Tue 11:30
Generation mix now
Gas60.0%
Nuclear14.7%
Biomass9.0%
Solar7.1%
Imports5.3%
Wind3.9%
Fig. 1Actual readings (solid) against NESO's own forecast (dashed), half-hourly. Forecasting problems like this are what we work on. Teal bars are zero-carbon sources. Updates every 30 minutes. Source: NESO Carbon Intensity API, CC BY 4.0.

Research-grade data science, built for real-world problems.

CountingLab was founded in 2010 as a spinout from the University of Reading. We take techniques from the research community and turn them into commercial systems that solve real problems.

Our work combines deep learning, heuristics and agent-based modelling, so each solution is shaped by the problem rather than by a favourite tool.

Methods

  • Forecasting
  • Clustering
  • Ensemble methods
  • Agent-based models
  • Optimisation
  • Large language models
  • Heuristics
  • Autoencoders & embeddings
  • Digital twins
  • Deep learning & visual AI
  • Social media analysis

Selected work

From national infrastructure to retail, across every kind of data.

  1. 2.1Deeper-ScanSecurity-screening technology developed with NWPro, bringing advanced computer vision to threat detection.Computer vision · Security
  2. 2.2Energy forecastingHousehold-level demand forecasting that preserves sharp usage peaks, enabling smarter battery storage and grid efficiency.Time series · Energy
  3. 2.3Community detectionCombining four community-detection algorithms to pick 18 meaningful communities out of 98,078 candidates in a large digital messaging dataset.Network science
  4. 2.4Understanding potential terrorist networksDynamic communicability on a network of 800,000 people and 4.5M communications: spotting influencers and emerging power by respecting the order of messages in time.Network science · Security
  5. 2.5Forecasting launch salesAgent-based modelling of the Coke Zero launch, forecasting first-four-week sales to within 18% error.Agent-based modelling · Retail
  6. 2.6Customer segmentationBehavioural segmentation with finite mixture models for a cashback website: over 40% of customers would have landed in a different segment under k-means.Clustering · Retail
  7. 2.7GDPR-compliant person re-IDPrivacy-preserving person re-identification. Full details for approved partners. Client accessComputer vision · Privacy
  8. 2.8CT scan analysisMedical-imaging analysis pipelines. Full write-up for approved partners. Client accessMedical imaging

Things you can try

Fig. 3.1sense.countinglab.co.uk

Synthetic data · Energy

SENSE-Synth

Synthetic energy time series for planners: research on generative models for demand data, turned into a tool an energy planner can use.

  • Generative AI
  • Time series
  • Energy

Published research

  1. [1]MEOW: Multi-Objective Evolutionary Weapon DetectionDaniel Dimanov, Colin Singleton, Shahin Rostami, Emili Balaguer-BallesterAutoML Conf 2023 (PMLR 224)
  2. [2]Estimating Network Dimension When the Spectrum StrugglesPeter Grindrod, Desmond J. Higham, Henry-Louis de KergorlayRoyal Society Open Science, 2024
  3. [3]Forecasting for Battery Storage: Choosing the Error MetricColin Singleton, Peter GrindrodEnergies, 2021
  4. [4]Diffusion Models for Generative Artificial Intelligence: An Introduction for Applied MathematiciansCatherine F. Higham, Desmond J. Higham, Peter GrindrodSIAM Review, 2025

Trusted by industry, government and academia

  • BAE Systems
  • Capgemini
  • DASA
  • Dstl
  • Energy Systems Catapult
  • Fugro
  • General Electric
  • Honeywell
  • Innovate UK
  • MOD
  • Net-a-Porter
  • Reckitt
  • SSEN
  • University of Oxford
  • University of Reading
  • University of Sheffield

From the lab

All notes
Bulgaria's Road Safety Challenge: A Call for ActionFig. 6.2

Bulgaria's Road Safety Challenge: A Call for Action

By leveraging advanced technology and realistic simulations, our prototype showcases our potential to enhance road safety with average speed checks, illustrating our commitment to innovative traffic monitoring solutions.

Have a hard problem?

We take on new projects with partners from industry, government and academia.

Start a project info@countinglab.co.uk