Data & AI

Matching candidate qualifications and experience to job requirements can be a laborious activity, let alone identifying where market skill gaps exist. To streamline this process we crafted a solution that leverages Gemini Pro as a Large Language Model (LLM) to generate matching scores to reveal how closely candidates’ qualifications match job requirements. Key features include PDF text extraction and real-time API processing. The web interface enables users to upload resumes and job descriptions, compare scores, and gain valuable insights into talent alignment.

Ensuring employees comply with protective gear requirements is critical for maintaining workplace safety, however monitoring compliance can be resource-intensive. To address this, we developed an AI-driven solution leveraging computer vision with OpenCV to detect whether employees are wearing helmets during machinery operation. The system incorporates machine learning algorithms, MySQL for data management, and Power BI for comprehensive reporting. By sending automated alerts to supervisors in cases of non-compliance, this tool improves safety measures and ensures effective oversight of protective equipment adherence.

Capturing key issues and resolutions from customer conversations is essential for improving support workflows but can be time-consuming and complex. To streamline this process, we developed a solution that transcribes and summarises audio and video conversations using Whisper, an advanced transcription model. Key features include seamless integration with a database and a user interface for efficient interaction. The system enables teams to quickly review essential conversation highlights, providing clear, actionable insights to enhance customer and employee experiences.

Ensuring data quality through cleansing and preprocessing can be a tedious and error-prone task. To simplify this workflow, we are developing an AI-driven tool that leverages natural language processing to make data cleaning more intuitive. Users can seamlessly upload Excel or CSV files, run natural language queries to clean and transform data, and download the cleaned datasets in real-time. Key features include AI-generated Python code using models like CodeLlama or Gemini 1.5 Pro, a user-friendly interface for efficient interaction, and support for tackling common challenges such as errors, inconsistencies, and data formatting issues. This solution optimises data quality management, making it accessible and scalable.

Providing responsive and accurate customer support in eCommerce can be challenging, especially when managing a high volume of queries. To address this, we developed an AI-powered chatbot using the OpenAI GPT-3.5 model, trained on FAQs to handle inquiries ranging from product details to order status. The chatbot includes features such as a vector-based query-matching system, human agent fallback via WhatsApp or an agent portal, and real-time notifications. This solution enhances the customer experience by enabling real-time interaction, resolving common queries efficiently, and escalating complex issues to human agents when needed.

The challenge of managing and monitoring health and safety risks across the diverse and operationally complex aviation sector requires a strong capacity to harness disparate data sources to convert into meaningful insights at both a governance and operational level. To deliver efficiency to this process we were able to ingest and categorise safety audit and incident reports using Microsoft Azure tooling and LLMs to streamline reporting requirements, and overlay RAG patterns to assist the auditing team in targeting priority areas of concern. This has delivered improved reporting outcomes, but also set the foundations for more adaptive auditing activity.

Managing fleet rental operations efficiently can be challenging due to fluctuating demand influenced by weather, time of day, and seasonal trends. To assess the potential of AI patterns to address this, we developed a machine learning solution leveraging XGBoost to forecast hourly demand for, in this case, mass hire bicycles. The system analyzes key factors and provides actionable insights for optimizing bike availability across the city. A web interface, built with Flask, enables users to input or upload data, generate forecasts, and explore multiple scenario predictions with real-time data validation. This solution empowers operations teams to efficiently allocate resources, ensuring bikes are available when and where they are needed to meet customer demand.