Tosin Dairo

• Built a rule based and machine learning Clinical Decision Support system that helps to diagnose early symptoms that leads to blindness with the adoption of a decision tree algorithm trained with best practice guideline data deployed on TensorFlow's Dialog Flow conversational bot. Demo can be found on https://dr-vee.herokuapp.com • Analysed and evaluated survey data generated from obstruction detector product tested in a school for visually impaired. • Deployed the NLP module for an open-sourced python library for beginners’ abstraction of data science and text pre-processing tasks, https://github.com/tosi-n/datasist/tree/master/datasist • Acquired Healtex feasibility study fund of £4,000 to extend Master's dissertation into a real life solution by releasing curated dictionary on veterinary medication. • Used medication NLP dictionary data to build deep neural network for named entity recognition, predictive compliance and usage tracking dashboard and a closed domain question answering model for drug active agents. • Built a deep spellchecker for syntactic and semantic corpus correction. • Also built an entity recogniser • Natural Language Processing Skills (Entity Recognition, Information Extraction, Relation Extraction, Question Answering, Conversational AI, Sentiment Analysis, Topic Modelling)

• Worked on auto insurance claims data to detect fraud in insurance claims. I used SMOTE oversampling technique on the minority class of the target class fraud reported which using the K-Nearest Neighbour algorithm to generate minority class. The overall accuracy of the Random Forest classifier model was 86.4%. • Worked on High Dimensional data comprising of gene expression data (2000 genes for 62 samples) from microarray experiments of Colon tissue samples. Used an Unsupervised Learning approach that used finding linear combinations of variables that capture most of the variability in the data and reduce dimension using Principal Component Analysis PCA. The ‘elbow’ of the scree plot was around the fifth to seventh principal component, which explained 90% variation.

• Built and deployed a dashboard for a webinar lecture I hosted to teach a community of beginners in data science learning R https://github.com/tosi-n/LAG-RUG-DASH • Built and deployed a clinical predictive model for predicting stroke in medical data https://github.com/tosi-n/Stroke-CPM

• Deployed an e-commerce database using MongoDB NoSQL database embedded in Django backend and Rest Framework in order to tunnel data and http request for an e-commerce website built on Angular frontend. https://tunneldb-api.herokuapp.com/api_documentation/