Jack Maughan

• Created On Demand Rankings (ODR), a league table that incentivize agents to contribute data. This creation of the this alone led to over 100% increase in contributions YOY as well as being regularly the most-viewed page on the news sight.
• Leverage Duplicate Detection / Anomaly Detection techniques to remove 95% of data point errors from the systems. Essential in living up to the EG motto of ‘The most trusted source of Real Estate Data”.
• Developed PropertyID a Network Graph with a Community Detection Algorithm that connected all disparate datasets together that automatically detected clusters that can be tagged as buildings. Currently being developed into the product but has already been sold as a stand-alone dataset for over £300k.
• Worked with Sales Operation team to develop Project Whitespace which utilized Third-Party dataset to find new customers. This also leveraged our own Salesforce data to create market segmentation and different pricing strategies and product bundling.

• Work alongside Legal SMEs to develop custom annotated datasets leveraging Prodigy.
• Set up the first product tracking tool (FullStory) which allowed us to track usage, redefine success metrics and perform A/B Testing with the UX team.
• Developed a custom Named Entity Recognition (NER) model that can be used as a Redactor tool, highlighting any information that can be considered PII. Obtained Recall scores of 92% and saved estimated time on this workflow by 80%
• Expanded a NER model to include Relationship Extraction for Citations and their sections within free text. This increased the eventual accuracy of Citation-Content matching by 40%
• Collaborated with Data Science teams and stakeholders globally to research the possibilities of LLMs and Generative AI. Collaborated with Anthropic to create Lexis+ AI with research focused on the areas of Hallucinations, Prompt Engineering and Grounding.
• Created a Sentence Classifier to Detect Judge Sentiment on Experts. Increased the F1-Score of this project by 20% to enrich our datasets on Context. Also expanded this scope to include positive/negative sentiment scores as a new product development.
• Present all of the above information in an easy to digest way, to all areas of the business from Sales to Directors so they are up-to-date with the current innovations.

• Japan LNG Model – Time Series Forecasting with an end-to-end solution. New Product development with high focus in Time Series, web-scraping and database management which helped drive up-sales to customers for end figures into the millions (£).
• Crop Disease Prediction ¬– Time Series modelling to help understand if we were able to predict when crop disease will occur and what preventative measure could be put in place to avoid this. This was eventually in an article in Farmers Weekly.