Harry Stamatoukos

Data Scientist across domains

As a data scientist, my work has been focused in 3 main areas:

• Algorithmic Pricing
• Forecasting
• Natural Language Processing

Specifically,

• On Algorithmic Pricing: I developed a model (XGboost) that decided which users should get a discount to optimize revenue. This model was deployed (VertexAI) via controled experiments and resulted in 3% increase in revenue per user.
• On Forecasting: I developed a model (Hierarchichal Forecasting) to forecast Active Riders, quarterly and yearly. This model was developed (VertexAI), an embeded with other forecasts while achieving an accuracy of ~2-5%.
• On Natural Language Processing: I used Large Language models (Google Gemini) to extract all information from users’ comments in a structured way that could be used by other systems. On top of that I built an automated report that summarised the key points of users’ weekly (GPT-4) which ended enabling the CPO and General Managers to drive weekly attention to problems.

Senior Data Analyst and Data Lead

In late 2021 I moved to Stockholm from London and started as a Senior Data Analyst before I became the team lead of 4. Voi, is one of the best and more complex data sets I have ever worked with, given the complexity of the micromobility industry. I was the first data hire of the rider domain. As a senior data analyst I took over the data ingestion airflow & python, the data modelling dbt , builiding dahsboards Tableau and geospatial analytics Unfolded & Python to drive the improvement of our core rider experience. As a team lead my team created the foundations for experimentation, advanced analytics (segmentation and LTV modelling) and proof of concepts on pricing and forecasting.

Data Science Coach

At Multiverse I was delivering lectures in Data Science and 1-to-1 coaching to 50 students. I delivered lectures in Statistics, Python, Machine Learning and general Data Science topics. Teaching is a very efficient way of learning fundamental topics very well, only when one has to explain something does one actually realises that one does not understand every layer of a concept. Over the course of 2 year, I delivered almost 30 lectures and 1200 1-to-1 coaching sessions.

Strategy and Data Analyst

At Colt I got my first job, where I was titled the “excel wizard”, primarily because I was good at googling. My first big achievement was an excell reporting automation, like a primitive BI, improving the way 40 service managers created reports - from spending a week to barely a few minutes. This enabled the company to increase the number of customers that coulde be service managed, increasing revenue by offering service management to more customers. After this achievement I was added to the leadership team reporting directly to the VP with a mission to drive more similar projects. This is where I started dabbling my data science journey, where I created a customer churn prediction prototype , KPI dashboards for the Executive Leadership Team and the data insights behind Colt’s digital transformation strategy. It was also the first time I got interested in the field of AI. Machine Learning was starting to become a buzzword.