Turning AI research into competitive advantage
We match the latest AI research to suitable use cases — working with companies to design, build and deploy cutting-edge Machine Learning models.
We focus on applying Artificial Intelligence to accelerate our clients’ core businesses.
An independent review of your existing AI system, with actionable recommendations.
AI for Executives
A crash course for your leadership team in AI, and what it means for your company.
For clients that are exploring Artificial Intelligence, here are examples of solutions that Neural Swarm can build.
Forecast sales and predict demand using historical time series data.
E.g. predict next month’s grocery store sales in order to optimise inventory levels.
Segment customers into groups by distinct characteristics.
E.g. optimise marketing efforts by assigning targeted campaigns to different customer segments.
Detect fake accounts and fraudulent transactions.
E.g. add additional checks for new account registrations that are deemed high-risk.
Solve logistics problems such as vehicle routing, scheduling and bin packing.
E.g. find the optimal placement of facilities to minimise transportation costs while avoiding competitor locations.
Detect abnormal conditions in industrial equipment using sensor data.
E.g. reduce unplanned equipment downtime by intelligently scheduling maintenance.
Personalised product recommendations and product rankings.
E.g. increase sales by recommending targeted products that the customer has never purchased before.
Text and voice conversations using an automated dialogue system.
E.g. minimise live agent interventions by automating inbound store switchboards.
Generate product descriptions, blog articles and marketing copy.
E.g. populate an e-commerce platform with fresh descriptions for every product.
Automate document capture and analysis at scale.
E.g. accelerate mortgage applications by automatically processing income and asset documents.
Determine whether a view on a subject is positive, negative or neutral.
E.g. monitor customer brand perception on multiple social media platforms.
Reduce long text passages into single-line summaries.
E.g. populate HTML meta description tags with a summary of the page’s content.
Real-time text and audio translation between languages.
E.g. offer instant translations of user-generated content to reach a global audience.
Detect inappropriate content within images and videos.
E.g. prevent toxicity in online communities by pro-actively moderating user-generated content.
Insert relevant ads at appropriate locations in videos.
E.g. monetise video content through advertising without adversely affecting user experience.
Verify identify or control access through facial recognition and detection.
E.g. determine if someone is wearing a face mask and whether it covers the nose and the mouth.
Automatically generate metadata for images and videos.
E.g. search a media library by audio track contents and on-screen objects and text.
Identify defects such as misplaced components or cosmetic damage.
E.g. improve quality control by detecting damaged components on a production line.
Match user-provided images with images from a product catalogue.
E.g. improve user experience by allowing users to search a product catalogue with a photo.
Explore some of our recent projects where we’ve made an outsized impact.
Horse Race Betting Model
Worked with a Hong Kong betting syndicate to model win, place, quinella and quinella place probabilities for horse races in Happy Valley and Sha Tin. The ensemble model applies Multinomial Probit with Maximum Likelihood Estimation (MLE) and other advanced techniques to past performance and sub-second money inflow data.
Package Delivery Time Simulation
Built a simulator for an Australian delivery company to assess business model viability. The simulator modelled the distribution of package delivery times in urban areas using uniformly generated point-to-point ETAs and ran over different times and days to account for seasonality. Deployed the project as an interactive tool comprising of a delivery time Isochrone overlaid over an urban map.
Increasing License Plate Recognition Accuracy
Helped an Australian vendor increase the accuracy of an existing license plate recognition (LPR) system by merging alternate data sources. The extra sensor data required real-time complex event processing (CEP) and unsupervised learning to extract a clean data stream to augment the current system.
Worker Productivity and Wellness
Developed and deployed a series of business intelligence dashboards that monitor factory worker productivity and wellness for a Hong Kong startup focusing on Corporate Social Responsbility (CSR). A series of reports, metrics and data visualisations were also researched and designed.
We frequently work with Cloud partners to deliver AI solutions that are highly available and scalable.
AWS Consulting Partner
Neural Swarm leverages pre-trained models in Amazon Comprehend and Amazon Rekognition to build baseline Machine Learning models. Amazon Sagemaker is used to train and deploy custom Machine Learning models.
Google Cloud Partner
Neural Swarm uses Google Cloud AutoML to build baseline Machine Learning models quickly and cost-effectively. Google Vertex AI is used to train and deploy custom Machine Learning models.
Get in Touch
Thank you for your interest in Neural Swarm. Please fill out the form below to enquire about our work, and we’ll respond as soon as possible.
Alternatively, schedule a free 30-minute introductory Zoom meeting.