Tayo – AWS, Python, PostgreSQL

cinch.co.uk is UK's largest online retailer of its own cars. The Vehicle Pricing Optimisation Model has been optimised to price 5000+ vehicles in stock until exit. The model is used to price vehicles daily with minimal human oversight.

• Built and operationalised a dynamic pricing engine (XGBoost Survival Embeddings, logistic regression, linear optimisation) using AWS Sagemaker, The model is used daily to price 5,000+ vehicles, improving annualised margin by £1.8M and accuracy by 5% vs. manual pricing. This model was selected in the top 5 in DataIQ's “Most effective stakeholder engagement solution of the Year 2024 category”.
• Managed senior stakeholder expectations with clear communication and empathy Collaborated with Data Analysts, Engineers, product Owners and Non technical audiences.
• Frequently presented the workings of technical models to C and D Level Executives with clear and concise language.

• cinch.co.uk is UK's largest online retailer of its own cars. With over 500k weekly visitors, its imperative to model user behaviour and use that to hyper-personalise and target customers via email and in real-time by generating a customer's propensity to perform their "Next Best action".

• Delivered a personalised email targeting model using LSTM on Adobe Analytics data
• Led the A/B testing to prove the value of the product to the marketing team, driving £800K in annualised margin uplift.
• Managed stakeholder expectations with clear communication and empathy.

Mealish.co.uk is a start-up looking to launch with an AI weekly meal planning tool. With access to scraped data from hundreds of restaurants in London, Mealish simplified the process of creating personalised weekly meal plans within seconds. It utilises Python, GCP, and GPT-4 to dynamically categorise a vast database of over 350,000 meals into a different dietary groups. This user-friendly tool streamlines meal selection, taking customers preferences into consideration and offers smart recommendations. It transforms meal planning into a convenient, diverse, and personalised experience, greatly improving the dining journey for users.

• Leveraged dbt for the transformation and cleaning of large restaurant meal datasets in GCP BigQuery
• Developed web apps using Python for backend services and data processing;
• Utilized GCP Compute Engine for scalable cloud resources;
• Deployed containerized applications using Docker for consistent environments;
• Automated CI/CD processes with GitHub Actions;
• Worked with GPT for advanced information processing and NLP tasks;
• Analyzed large datasets using Pandas and Scikit-learn for data wrangling and machine learning;
• Managed datasets with BigQuery to support decision-making;
• Conducted API testing for reliable functionality;
• Seamlessly integrated Stripe API for payment transactions;
• Classified over 350,000 restaurant meals into dietary plans like vegan, gluten-free, and halal;
• Generated over 5000 meal plan combinations tailored to user preferences;
• Streamlined the merchant onboarding process using A/B tests for quicker sign-ups;
• Developed dine-in REST API and meal plan REST API with a FASTAPI microservice architecture;
• Created a Streamlit app prototype to gather initial feedback on app functionality from stakeholders.

Created a spatial map of inferred pipe assets (private drains and sewers) using Machine Learning to support Severn Trent Water Engineers in planning, maintaining, and repairing leaks and bursts within the Severn Trent Water network.

• Generated one of the UK’s largest inferred sewer asset bases spanning 34,000 kilometers by performing big data analysis on millions of sewer assets using a combination of on-premises task scheduling infrastructure and GIS tools, including QGIS, PySpark, Python GIS libraries, and Postgres database in AWS;
• Utilized regression and classification algorithms such as XGBoost, random forest, and simulations using AWS Sagemaker notebooks to estimate the length, location, shape, and diameter of sewers;
• This saved Severn Trent Water hundreds of resource hours and millions of pounds required for manually surveying unmapped assets.

Optimising and reporting on marketing campaigns for 65 clients spanning across Ukraine, Ghana, Kenya, and Nigeria.

• Implemented, monitored, and generated over 200 ad-campaign reports for 60+ direct clients and ad agencies, alongside actionable insights to improve Return on investment;
• applied regression models in Python to forecast campaign performance and identify areas for improvement;
• Developed and applied Python-based regression models (using scikit-learn) to forecast 2016-2018 financial year revenues, aiding strategic investor discussions and identifying key revenue drivers;
• Managed data storage and access using Microsoft SQL and ensured data quality and security;
• Utilised research, survey analysis, insights from social media, Google Analytics, and data science techniques to analyse customer behavior, market trends, and campaign effectiveness. Used this to generate reports and media kits to support sales managers.