Chris Ayyildiz | Portfolio

Data Analytics Projects

8 Projects

Live Funnel Analysis

Simulates live e-commerce user activity through an interactive funnel to monitor, visualise and improve user conversion pathways. The project features a real-time ETL pipeline and an interactive dashboard to identify potential drop-off points.

Python Pandas Streamlit Plotly

View Project GitHub

Access Transport

Automated ETL pipeline identifying live TfL accessibility disruptions via their unified API. Ingested data is processed with Airflow and outputted to a Streamlit dashboard enabling users to find relevant real-time accessibility issues.

Airflow Asyncio REST API Streamlit

View Project GitHub

VISA Climate Hackathon

Worked with a 500k+ row dataset of anonymised transactional data to identify trends in fast-fashion spending. I mapped spending to carbon footprint estimations using gov.uk conversion factors and resolved significant mismatches in column entries.

Jupyter Notebooks Python Pandas Plotly

View Project

Web User Behaviour Analysis

Simulated 100K+ web sessions using Python and Faker to analyse web behaviour, identifying patterns in engagement, session metrics, and conversion. Fully containerised for rapid reuse.

Python Streamlit Plotly Docker

View Project GitHub

NYC Taxi Demand Analytics

This project analysed over 1.2 billion NYC Yellow and Green Taxi trip records to practice using large scale cloud tools for data analysis. Key outcomes for the project included a peak demand heatmap spaced across time periods, revenue per mile trends throughout the past five years, pick up cluster spots, and trip duration anomalies. The project used the GCP as part of Google's Machine Learning Engineer Pathway.

BigQuery Cloud Storage dbt Looker Studio

KPI Tree (NHS Metrics)

Real-time NHS provisional care quality KPI board to track performance status. Utilises web-scraped data and recent public databases to collate KPI data in a React dashboard.

Beautiful Soup PostgreSQL React ETL

View Project

UK House Price Change (1995-2024)

Visualises data from the England & Wales Land Registry Price Paid dataset on a regional scatter heatmap across 2024 and 1995.

Python GeoPandas Plotly Mapbox

View Project

Global Urban Heat Island Visualisation

This project visualised the significance of the Urban Heat Island (UHI) effect across global cities over the past three decades. I used LST data and urban boundary shapefiles to generate an clear heatmap graph comparing cities to their rural surroundings.

Google Earth Engine Kepler.gl QGIS GeoPandas

Machine Learning Projects

4 Projects

Air Quality Forecasting

Deployed an air quality forecasting tool taking API data from OpenAQ and conducting time series analyses using Prophet to identify seasonal pollution spikes. The project collected over 14,000 hourly PM2.5 and PM10 data readings.

Prophet PyTorch REST API Docker

Carbon Emissions Forecaster

Built and validated a supervised regression model which forecasts U.S. carbon emissions using EPA datasets as part of the Google Machine Learning Engineer Pathway. The project achieved an R2 = 0.89, RMSE = 0.13.

GCP BigQuery Time-Series ML

Adaptive Pricing Model

Developed a revenue maximising pricing system taking competitor pricing, device identifiers for user segmentation, and demand context (e.g., route popularity) which outputs a price. The project is optimised as a FastAPI microservice offering low latency responses even under load.

FastAPI Redis Pydantic Pythin

Green Energy Loading Scheduler

Deployed a scalable ML based enrgy scheduling platform for UK commercial buildings using probabilistic time-series models to forecast weekly energy demand. The system generates optimal schedules reducing carbon intensity by approximately 26%. Ready to be integrated with BACnet and Modbus.

FastAPI Streamlit Prophet OR-Tools