I turn messy data into models that actually ship.

2024 – Present

United Kingdom

← where I am now

Work Coach, Department for Work and Pensions

• –Communicate complex benefit entitlements and administrative processes clearly to claimants from diverse backgrounds, adapting language and explanations to individual needs.
• –Built a personal Excel tracking system to manage a high-volume caseload, enabling data-driven prioritisation, deadline management, and early identification of at-risk cases.
• –Collaborate with employers and training providers to match claimants with suitable opportunities through effective relationship management.

Excel · Stakeholder communication · Case management

Jan 2024 – Oct 2024

Remote, UK

Freelance Python Developer, AscentiQ

• –Developed a customised trading application for a financial-services client using Python and PyQt6, enabling real-time trade execution, portfolio tracking, and interactive visualisations.
• –Built an end-to-end data pipeline on Google Cloud Platform that gathered, scraped, and processed financial data from multiple sources; containerised with Docker to fully automate cloud deployment.
• –Created interactive dashboards and financial reports using Plotly Dash and PyQt6, translating complex financial data into actionable visualisations for client stakeholders.

Python · PyQt6 · GCP · Docker · Plotly Dash · Web Scraping

Jan 2022 – Nov 2023

Weybridge, UK

Data Scientist, AXA UK

• –Collaborated with Accenture on a contact-intent and channel-switch optimisation project, gathering unstructured raw data from GCP and applying data-engineering methods to enrich datasets for ML modelling.
• –Identified customer trends through EDA and automated stakeholder reporting, improving the digital self-service journey, giving a 30–60% reduction in calls and an estimated £0.4–0.8m business benefit.
• –Built a custom LSTM and fine-tuned a Facebook Prophet model to predict call-centre volumes, improving forecasting accuracy from 813.35 to 527.1 RMSE.
• –Developed and deployed a custom TensorFlow deep neural network using PySpark to classify complaint outcomes, with Optuna hyperparameter tuning and MLflow introduced to track experiments across the team.
• –Proposed and implemented transformer-based summarisation with Hugging Face, benchmarking BERT and T5-small and fine-tuning T5-small for production, which significantly reduced agent review time.
• –Led daily stand-ups and sprint planning, coordinating three colleagues across teams to deliver a Plotly dashboard surfacing classification and summarisation outputs.
• –Analysed Feefo and Trustpilot reviews with N-grams and NLTK to identify complaint themes and price ranges driving negative feedback, building a customer-satisfaction monitoring system.
• –Implemented a random-forest regression model to predict compensation amounts, delivering 15–20% cost savings.

Python · GCP · PySpark · TensorFlow · LSTM · Prophet · Hugging Face · BERT · T5 · Optuna · MLflow · NLTK · Plotly

Aug 2021 – Dec 2021

London, UK

Data Engineer, Internet Bull Report

• –Built a Python application using Beautiful Soup to scrape financial data from FinViz and Yahoo Finance, cleaning text and creating structured datasets for each penny stock by market sector.
• –Developed LSTM models to predict stock prices and ran sentiment analysis on financial news headlines (positive, negative, neutral).
• –Implemented the entire pipeline end-to-end in GCP, fully automating data collection, modelling, and reporting.
• –Achieved an 80% increase in weekly report subscription rates by delivering more accurate, data-driven insights.

Python · Beautiful Soup · LSTM · TensorFlow · NLP · GCP