Part of the cast and crew for several scenes in "You Me Bum Bum Train"

Current Contract:

• Building a full web-based application to tackle the company’s specific needs: Right now their ERP system is Sage200 which can’t handle multiple units of measures for procurement effectively, can’t adapt to the fluctuating pricing of raw material pricing, and doesn’t align with the companies current data processes. As such, I’ve been tasked to build a web-based ERP system which tackles all these problems

• Because of the specific problems outlined, the application is tailor made based on what the current managers want.

These include aligning with their current data processes, allowing users of the company to have more control over the

software features, and adding unique features that’s related to food industry QA testing

• React/Node.js for the frontend and Django for backend

• Created python GUI applications to parse and automate much of the procurement and HR processes in the company (From sales orders to purchasing)

Joint Research dissertation: Predicting solar flares with Machine Learning [https://github.com/aranan101/met-office-project]

• Parsed and cleaned twenty years of raw data on Solar flares from three different sources using the SunPy and drms package. Merged the raw datasets into concise data for Machine Learning modelling. Also engineering the inclusion of Mcintosh Evolutions into the data, a novel feature in the classification of solar flares.

• Trained Support Vector Machine models (SVMs) to forecast the strength of a future solar flare given data 24 hours prior. Hyperparameters were tuned using cross-validation and multiple runs with different combinations for each parameter.

• Implemented feature selection models to reduce the dimensionality of the datasets such as F-scoring, Logistic regressions, and Linear SVM with l1 regularizer.

• Created novel modifications to the SVM model by merging it with the KNN algorithm, bagging techniques, and the SMOTE algorithm

• Matched the Met Offices industry-standard forecasting method in terms of prediction capability (measured by the ROC AUC. The final ROC AUC during testing was 0.89 ).