Technology is best judged in context.
This selection of our AI engineering projects shows how we turn real business and operational challenges into working systems that reach production.
Client's name and type: u-blox (technology solution provider)
Industry: Information Technology
Project/Solution: Deep learning-based indoor positioning using Angle of Arrival (AoA) estimation
Problem:
Indoor positioning is in high demand for uses such as navigation in hospitals, offices and airports, and locating objects in warehouses. Bluetooth Low Energy (BLE) is an affordable, energy-efficient option, but its accuracy drops because signals reflect off walls and objects. u-blox wanted to explore how deep neural networks could make BLE-based positioning more accurate and robust.
Overview:
We developed deep learning models that estimate the position of a BLE tag from the signals received at multiple anchor points. Using the IQ components and signal strength (RSSI) of the received signals, the models estimate the Angle of Arrival at all anchors at once, and a least-squares positioning engine then calculates the tag’s position. Some architectures distribute the processing across the anchors, which makes embedded implementations easier.
Tools developed:
· Deep learning architectures for AoA estimation, including multilayer perceptrons and CNNs
· Joint multi-anchor AoA estimation with distributed processing
· Least-squares positioning engine
· Data augmentation strategy that reduces training data needs
· Open high-resolution dataset from realistic ray-tracing simulations
Impact:
The models achieved 70 cm localisation accuracy and stayed robust when furniture, obstacles or anchor layouts changed. The work was published in the peer-reviewed journal Sensors (MDPI), and the dataset and models were released openly to support reproducible research.
Challenges met: Signal reflections and multipath propagation, robustness to changes in the environment, computational load on embedded hardware and limited training data.
Technologies used: TensorFlow, MATLAB and ray-tracing simulations.
Client's name and type: Piraeus Container Terminal (PCT)
Industry: Ports and Logistics
Project: 5G and AI-enabled driver fatigue monitoring system
Tools developed: Computer Vision architecture, Far Edge Module, Near Edge Module, Business Intelligence Framework, interactive augmented dashboards, Data Warehouse.
LIBRA AI's RESONATE is an advanced driver fatigue monitoring system that addresses the pressing issue of driver drowsiness, a significant contributor to motor vehicle accidents. Developed and extensively tested at the Piraeus Container Terminal, RESONATE prioritizes the safety of professional drivers, mainly those operating heavy machinery and trucks. RESONATE employs continuous video-feed monitoring of the driver's face and eyes to assess an individual's alertness level automatically. The system adopts a hybrid approach, seamlessly integrating far-edge (in-truck) and near-edge (server) submodules, harnessing the power of 5G technology for real-time drowsiness detection. When signs of drowsiness are detected, RESONATE responds immediately by triggering an alert that helps prevent potential accidents. In addition, RESONATE incorporates a robust business intelligence framework empowering organizations to extract valuable insights from fleet behaviour, facilitating data-driven decisions and significantly enhancing overall safety within their heavy machinery fleets. The system is GDPR compliant as the video stream is processed on-the-fly.
Challenges met: Lighting conditions; Intensive high and mid-frequency vibrations; Driver seat suspension; Guarantee a low number of false alarms; GPU-enabled edge devices supply limitations; Privacy concerns;
Technologies used: Kafka, Flink, Tensor Flow, Edge Computing
Client's name and type: Coffee Island (retailer with more than 400 franchise shops)
Industry: Food and beverage
Project: Ecommerce data analysis and Customer Segmentation based on Sales Data
Tools developed: Data warehouse design for collection and analysis of all the e-commerce data of the chain. We built a customer segmentation framework (by RFM analysis) based on the customers' buying behavior dedicated to inspection and direct marketing campaigns. We also created an e-Commerce sales dashboard set to enable a 360o prompted inspection and performance monitoring of the chain's stores and products.
Challenges met: Handling a vast amount of data from several different resources to develop a tool that helps CI team to build its chain's strategic roadmap.
Technologies used: Apache NiFi, MySQL, Postgresql, Python, Pandas, Tableau Online, RFM analysis
Client's name and type: BNDB & Lambda developments
Industry: Commercial Real Estate
Project/Solution: All Marketing Analytics in one dashboard suite.
Problem: Marketing data across Golden Hall, The Mall Athens, Mediterranean Cosmos, and Designer Outlet Athens was fragmented across websites, campaigns, and marketing technology platforms. Teams lacked a unified way to monitor key marketing KPIs, understand audiences, evaluate campaign performance and optimise marketing investment.
Overview:
We developed a Tableau Cloud dashboard suite that centralises marketing and customer data across LAMDA Development’s shopping centres. It provides a consistent reporting environment for campaign monitoring, audience analysis and performance evaluation.
Tools developed:
Impact:
The suite gives stakeholders a single view of marketing KPIs, customer insights, and channel performance, enabling faster decision-making, sharper targeting, better allocation of marketing investment, and improved ROI.
Challenges met: Data inconsistency, multi-source integration, high-volume processing, ETL design, Tableau performance, cloud deployment, multichannel attribution and marketing data monitoring.
Technologies used: Tableau Cloud, Tableau Desktop, MySQL and Python
Client's name and type: National Bank of Greece (Financial Institution)
Industry: Finance
Project: Machine Learning / Natural language processing competition (Link)
Tools developed: LIBRA co-developed and supported the NBG’s Information retrieval in court decisions challenge.
This included the development of a hybrid pseudo anonymizer to remove sensitive data from the court decisions (OCR scanned documents in the Greek Language). The anonymizer was a hybrid solution, utilizing NLP techniques to identify sensitive entities (such as NER and POS), as well as pattern matching and term frequency related techniques to achieve best performance.
Additionally, we:
Challenges met: Automatic question/answering in court decisions based on cutting-edge NLP models. Information Extraction in the form of passage retrieval. Annotation of very large documents requiring domain-specific expertise.
Technologies used:NLP and Deep Learning frameworks such as spaCy, nltk, gensim, TensorFlow, PyTorch.
Client's name and type: Greca (Tourist Agency)
Industry: Tourism
Project: Personalized Product ranking Recommendation System supported by advanced A/B testing
Tools developed:
We developed a Recommendation System for our client's products driven by web behavioral data. The system includes various recommendation engines for the different sections of our client's website, such as product re-ranking recommendations. To develop the recommendation system, we started off with data analysis to determine parameters like visit duration and product inspection journeys, in order to design preference profiles among products and traveling destinations that were used to enrich the personalized recommendations we generated. The system provides recommendations both for new and returning visitors using different strategies. In the whole architecture, we had to develop a solution to store log data of recommendations that would provide the required data integrity to perform advanced A/B test experiments that validate the recommendation strategies. Finally, we developed an API-based end-point to provide recommendations to our client whenever a visitor visited a web page of interest.
Challenges met:
As a third party providing this solution, we had to develop a series of tests to validate the successful integration of the system, assuring the quality of user experience. This process included heavily analyzing web behavioral data and monitoring in real-time the performance of the API we deployed. Keeping track of product changes was also a major challenge requiring regular recommendation updates.
Technologies used: Apache NiFi, Μatomo, Python, NodeJS, PostgreSQL.
Client's name and type: Greca (Tourist Agency)
Industry: Tourism
Project: MatTech analytics framework with Algorithmic Attribution Modeling & Anomaly Detection
Tools developed:
We developed a Data warehouse gathering the e-commerce data, web behavioural data and digital campaigns data. We delivered a custom marketing analytics Business Intelligence dashboard suite that supports advanced algorithmic attribution modelling and anomaly detection for Ads spending optimization.
Challenges met: To thoroughly monitor e-commerce fused with web behavioural data along the full customer journey across all touchpoints and perform impactful marketing campaigns.
Technologies used: Apache NiFi, MongoDB, Postgresql, Python, Pandas, Tableau Online.
Client's name and type: PNOE (Start-Up)
Industry: Health Care / Fitness
Project Type: Streaming processing and analytics, on-the-fly ETL, scalability on-demand, Fast OLAP Database
Solutions developed: PNOE project required real-time processing, transformation and analytics of big multi-user streaming biometric data. The aim is to provide real-time analytics and demanding algorithmic calculations on several health/fitness metrics displayed on PNOE mobile app. Such requirements cannot be covered with old fashioned data warehousing, i.e. you collect all data and perform the aggregations and data transformations overnight. Our solution stack was designed to allow on-the-fly, true stemming data processing through Kafka and Flink and to seamlessly support scaling, both in terms of the evolving app complexity and the growing user base via Kubernetes orchestration in AWS.
Challenges met: The task involved both real-time streaming data as well as out-off time, retrospective streams. This is a challenging situation for and Flink's functionality had to be extended to support both data modalities in a common pipeline. Also, the ETL has been fine-tuned to prevent the memory/processing load from growing out of proportion.
Technologies used: Apache Kafka; Apache Flink; Scala; Kubernetes.
Client's name and type: Alpha Marine (Marine Consultancy Firm)
Industry: Maritime
Project: Visualization engine for the company's Deep Dive Project and causality/regression analysis.
Tools developed: We developed bespoke visual analytics and data science studies for the Alpha Marine's Deep Dive Projection Shipping Companies' Culture. A major deliverable was the Deep Dive data Parser, which takes care of the data retrieval stage. Special care has been given for the Deep Dive data Parser to tackle all those inconsistencies among the surveys in the most generalizable way. The static visual analytics approach followed, where an extensive set of diverging infographics have been designed and programmed in Python to visualize the information and get insights inherent in the data related to the data produced through the Deep Dive surveys. We also provided automated trend analysis and sparse regression analysis tools and approaches.
Challenges met: Before any analysis, the data must be retrieved, prepared, and homogenized in a shape that is liable for analytics In the Deep Dive case. This poses particular challenges, mainly due to the multimodality of survey response tools (both survey monkey and custom Excel worksheet).
Technologies used: Python, Pandas, Matplotlib, Scikit, Tableau