Phil Marsh Email & Phone Number

London, Gb

Building the backend API for a new Baseball mobile game. We want to ask interesting questions, relevant to the live game the player is watching, and reward correct answers. To do this, we need to be able to predict the course of the game, and turn those predictions into questions with answer choices.Skills: System Architecture · Machine Learning Engineering · Data Engineering · Python (Programming Language)

London, England, United Kingdom

Tackling cheating in online games utilising the power of Machine Learning to scale review to your data volumes. We manage the complexity of the data pipeline, maximising convenience for game developers.

London, England, United Kingdom

- Designed and built a behavioural customer segmentation product, intelligently handling seasonality without ignoring genuine behaviour change.- Modified and integrated anomaly detection software at a Tier 1 Bank.

Milano

- Built a Gated Recurrent Unit Recurrent Neural Net (GRU-RNN) to make product recommendations using website clickstream data. (python/keras)- Prototyped using deep reinforcement learning to automatically adjust product pricing. (python/gym/tensorflow/keras/statsmodels)

- Built an MLP Neural Net to make product recommendations for flight ancillaries. (python/tensorflow)- Led a project to prototype an approach to personalise product imagery using a Convolutional Neural Network (CNN) (python/keras)- Ran training for Group data scientists on building Convolutional Neural Networks (CNNs) (python/keras)

Projects:- Deloitte/Leading Investment Manager in the UK (London - 12 months) - Implementing trading compliance and market abuse monitoring system backend, including trader profiling and anomalous trading activity detection. (python/pandas/sql/sklearn)- Datagenic (London - 5 months) - Built a news collection and sentiment analysis engine for commodities, offered as a product to Datagenic's customers. (python/sklearn/spark/kafka/mongodb)- Deloitte/Major UK Retail Bank (London - 6 months) - Built a system to extra commercial customer's financial data from OCR output of their financial and management accounts. (python/pandas/sql)

- Built a model to predict PPI claims volume and monetary cost.- Outlier detection to detect unusual behaviour for cyber security analysts to investigate.- Built Lifetime Expected Loss (LEL) model to quantify mortgage risk.