We employ artificial intelligence and machine learning techniques to build data-driven platforms that stand out in personalisation, optimisation and innovation. Industries we have served include: financial services and insurance, transport and logistics, retail, telcos, and oil&gas.
Sample case studies below:
Smart City End-to-End Platform
ADDO AI led a highly strategic and innovative project on mobility-as-a-service for one of Asia’s largest public transportation providers. Our engineers and AI experts are developing an AI-based urban mobility solution that would allow residents to search, book and pay for any multi-modal journey through a single user-friendly smartphone application. The complex platform requires integration with many different mobility providers including bike-sharing, buses, trains, e-scooters and autonomous vehicles, and machine learning enables the platform to provide personalised mobility recommendations to commuters, and to recommend infrastructure changes in response to demand prediction.
Data Lake and AI Use Cases
ADDO AI is working with one of the largest airlines in Asia to build a scalable and serverless data Data Warehouse (DW) platform, which can provide provide in-depth data exploration and visualization of customer data using integrated BI tools. The system being developed by ADDO AI presents holistic data governance support where the infrastructure is capable of handling terabytes of data. With our data lake and cloud architect specialists, we are able to deliver serverless, integrated, and end-to-end data warehouse services that provide scale, performance, security, and unmatched cost efficiency.
Recommendation Engine and 360° View for Retail Shoppers
ADDO AI has developed recommendation engines for multiple clients including one of the world’s largest credit card company and a leading retail chain. We build recommendation engines are built using a range of techniques including collaborative filtering, matrix factorisation, content based filtering and neural network based recommendations. Data at the SKU level is analysed real-time and then recommendations are fed into marketing campaigns, and customer journey interventions. One project resulted in a significant gross sales increase from the recommendation engine.
Customer Service Automation Using Conversational Chatbots
ADDO AI has designed and implemented several conversational chatbots for the logistics and telco industry. Using natural language processing and integration with existing customer relationship management systems, the chatbot is able to automate, personalise and optimise customer interactions. Measurable KPIs have included reduction in customer waiting time, increased net promoter score in customer satisfaction and reduction in staffing costs. Chatbots were built using existing engines such as Google’s DialogFlow augmented with our own custom intent classification to create more conversational flows. Languages included Mandarin and English.
Anomaly Detection for Payment Fraud
ADDO AI is building a framework for a top Asian financial services company to automate the process of abnormal activity detection in financial transactions. Our framework detects outliers based on domain specific rules and machine learning techniques like Bayesian classification for abnormal activity risk classification. Our solution generates user transaction profiles by obtaining user behaviour according to historic transactions based on categorical or numerical attributes, and then triggers flags based on unusual activities.
Automated Operations and Incident Management
ADDO AI is working with one of the world’s largest telecommunications firms to IT to develop applications and AI engines that disrupt their classical delivery model of incident management and enterprise customer service. After integrating with their core operations systems, ADDO AI uses a combination of artificial intelligence techniques (support vector machines, gradient boosting trees and deep learning) to automatically classify incidents from different communication channels, conduct root cause analysis and trigger actions for operational support. This immediately lowered cost of operations management, improved SLAs and human error for our client.
Predictive Maintenance and Input/Output Optimisation with IoT
ADDO AI developed a predictive maintenance engine for a leading Oil & Gas organisation using hundreds of sensors in manufacturing machines and compressors. We have also developed impact analysis for a leading fertiliser company in South Asia to optimise the input/output in order to increase revenue and production. In each case, ADDO AI built and trained a proprietary model using sensor data fusion techniques to achieve these goals, and implemented a distributed data architecture to enable real-time predictive analytics.
Risky Driving Prediction for Insurance Using Computer Vision
For one of the largest insurance providers in Asia, ADDO AI built a prototype platform that identified the risky driving behaviour of elderly drivers. Using a combination of data from car cameras, telematics from the on-board diagnostics (OBD) and the traffic and weather information, the model was trained to identify and predict when elderly drivers begin to drive in a risky manner so that they can be advised to stop and park their cars. The pilot used a variety of AI techniques including deep learning and convolutional neural networks, and an important aspect was sensor data fusion and handling.
Data Lake Architecture and Data Migration
ADDO AI has helped several Enterprise clients move their data from siloed and disparate databases into a central data lake that can be used for firm-wide business analytics and machine learning. Data lakes provide more agility and flexibility than traditional data management systems. Our data lake architects help clients (including one of the largest banks in Asia) determine their overall end-to-end data architecture and analytics strategy. This can include any combination of open-source, cloud and on-premise vendor solutions. Then our data architects help extract, transform and load data into the data lake, making it ready for our AI experts to develop insights and trigger actions. Data Lakes can be built in an iterative manner over the course of several months.
Resource Optimisation using Forecasting
ADDO AI worked with one of the largest US hospitals on a resource management project to optimise hospital staffing problem based on predicting patient demand. By collecting, analysing and preprocessing resource occupancy data, staffing predictions are generation based on techniques like Bayesian tIme series analysis, Facebook prophet, Lesso regression and Vector autoregressive model are used for predictions.
Disease Outbreak Prediction
ADDO AI worked on a disease surveillance system. This involved the collection and analysis of real time information such as weather data from Meteorological Department, disease outbreak data from Health Department, and health care data from hospitals and threshold constants values from domain experts across the Punjab province. Prediction patterns were published in ICDMW and underscored the influence of variables such as literacy rate on health outcomes.
Smart Building Optimisation Strategy
ADDO AI was a strategic advisor to one of Singapore's leading real-estate startups to use technology and data analytics to support its regional expansion across Asia. Optimisation strategies include maximising revenue per square foot by analysing occupancy using historical data, iOT, and contextual information and training a neural network to predict occupancy patterns and develop strategies to increase revenue. The strategy relies on converting every part of the real-estate, operations processes and client behaviour as data points that can be used to personalise offerings on the internal services platform, offering customers seamless services when and where they need it.
Data Governance and Identifying Bias in Algorithms
The ADDO AI leadership devised a constraint based novel data mining model to undermine the social discrimination against certain sensitive groups, such as females, blacks and minorities within society. The model was put forth for unbiased policy making by the law enforcement agencies. The developed techniques are deployed in The Dutch Department of Justice (WODC) and The Dutch Central Bureau of Statistics (CBS). His proposed solution was taken up by the international data mining community (including from Microsoft USA, Harvard University, University of Toronto, Kyoto University and University of Pisa) who are building upon the research.
Personalisation for Citizen Services
ADDO AI was selected to work with Smart Dubai and Dubai Futures Accelerator to develop an AI-powered platform that improves the speed and quality of citizen services by personalising services and anticipating citizen services and needs. The platform uses a mixture of collaborative filtering, deep learning and adaptive learning to enhance the accuracy of personalised recommendations.