Oliver Blane - Portfolio

Who am I?

I'm Oli, a Physicist disguised as a Data Scientist. Feel free to skip the pleasantries and take a look at my projects.

I have 3 years of professional experience in Data Science, currently leading a team of two other Data Scientists. I graduated with a First Class Honours degree in Physics from Queen Mary University of London, and was awarded Principal's Prize for outstanding academic achievement.

I have worked on various projects ranging from Generative AI, to Data Analysis and Visualisation, Conversational AI using Natural Language Processing, and much more.

This portfolio contains some of my favourite projects that I have worked on throughout my professional career and my studies.

Key Technical Skills

Python (Pandas, Scikit-Learn, NumPy, Matplotlib, Keras)

Artificial Intelligence

Cloud Computing (Microsoft Azure, AWS)

SQL

Power BI

My Favourite Projects

Estimating a daily R_t value for COVID-19

When COVID-19 began to sweep across the UK, there was huge emphasis on the R number. Government officials insisted that we must get R below 1, and keep it there. No exact values were ever broadcasted, only wide ranges in which R may lie, and so I decided try and estimate R (within error) for each day. The figure on the left shows some of my key findings for the UK up until the middle of May 2020 (see my GitLab for more).

Languages: Python (Pandas, NumPy).

Rapid generation of courses tailored to a specific individual

Generate courses for any topic at any level in any language within minutes.

The tool uses a variety of large language models to create course outlines, lesson content, quizzes and interactive content, along with translation models to convert the material into any language, and voice synthesis models to bring the content to life with audio. Two distinct image generation models are used to generate lesson images, depending on the type of content.

Languages: Python (OpenAI, Anthropic, Microsoft Azure AI Speech Services, AWS).

A virtual assistant for students

I built a chatbot to make life easier for students, for example enabling them to change their enrolled courses, find out their teachers for each course, and book mental health appointments with their college's mental health professional.

Languages: Python (using the RASA framework).

Machine learning the properties of perovskite structures

Perovskites are chemical compounds that are used in fuel cells and electrochemical sensing. Knowing their properties is very important for this reason, and thus I built a dataset of over 100 perovskites with over 30 features for each and used classification and regression models to predict these features.

Languages: Python (Scikit-Learn, Pandas, NumPy).

Predicting the stability of compounds using deep learning

As an extension of the previous project, I utilised artificial neural networks with a dataset of 2 million compounds to predict the formation energy, which is a property that dictates compound stability. I compared various neural network structures to find which performed best.

Languages: Python (Keras, Pandas, NumPy).

Track - building a mobile workout tracker that focuses on data analysis

Fitness is a passion of mine, and so I decided to start developing a mobile app that meets my workout needs. The long term plan for this fitness tracker is to accomodate those who want the most detailed analysis of their workouts, for example by importing AppleWatch workout data. This is an ongoing project that I work on during my spare time.

Languages: JavaScript and React Native for front-end, Python for data analysis.

Analysing Sleep Data

Sleep is one of the most important factors in living a healthy and happy life, and I find it extremely interesting. I tracked some key metrics describing sleep quality, along with many variables that I hypothesised could influence quality of sleep.

For example, the interactive graph shows how temperature affects my sleep.

Languages: Python.
Tools: Power BI, Excel.

Analysing Lego data from the last 70 years

Ever wondered how the number of pieces of Lego per set has changed over the last 70 years? I did, and so I used some Lego data to find the answer (along with the answer to some other questions...).

Languages: Python.

Transition temperature of dielectrics using quasielastic neutron scattering

Dielectric materials are used in all sorts of electrical components. I investigated at which temperature the behaviour of these dielectrics change, and what effect this has on the energy barrier to molecules within the dielectrics.

Languages: Python.