Portfolio

In this notebook we will demonstrate a simple forecasting technique using data stored in Clickhouse. We will use a dataset regarding the number of passengers departing an airport for different airlines, and aim to predict this time series into the future.

In this notebook we will demonstrate a simple time series anomaly or outlier detection technique using Clickhouse. We will use a financial time series, specifically a daily time series of the BTC-USD and ETH-USD exchange rates between 01 July 2019 to 30 June 2020, aiming to identify outliers in this period.

In this notebook, we analyse the risk of a portfolio of stocks using only ClickHouse functions. We use the historical time series of 120 stocks included in 4 different indices. We calculate a number of risk metrics including the widely used "value at risk" measure.

In this notebook we will demonstrate a simple linear regression example using Clickhouse Stochastic Linear Regression function. We will be working with a last mile delivery dataset, and will attempt to predict how long deliveries should take as a function of the delivery distance and pickup time. We will make use of Clickhouse's geoDistance function to work with geographical data and render geographical data as part of this Hex notebook.

In this video we introduce AWS Bedrock, the new AWS feature for managing and using large language models (LLMs). We demonstrate how it can be integrated with proprietary knowledge bases (RAG) and how we can build agents with no code.

In this video we demonstrate how to integrate dbt and Clickhouse Cloud. The video includes an introduction to dbt, and a discussion as to whether we should integrate dbt and Clickhouse in this way.

ClickHouse Cloud and Hex are a lightweight, pragmatic, fully managed stack for analytics and data science based on real-time data. In this video I demonstrate the two technologies working together and how they can be used to implement a develop process and build low code apps.

In this notebook we build a simple recommender system based on collaborative filtering using only Clickhouse functions.

In this notebook we demonstrate an optimisation problem in the logistics industry, whereby we have to load various vehicles according to constraints and then plan an optimal route for their deliveries. This is a variation of the travelling salesman problem.

In this notebook we will demonstrate a simple time series classification technique using Clickhouse machine learning functions. We will use a dataset describing power demand in Italy, with a classification task to distinguish days in the fall and the winter.

In this video we demonstrate how large language models can be used within the financial services industry for knowledge discovery and automating document production. Technologies used include ClickHouse and AWS Bedrock.

In this notebook we will demonstrate a simple workforce optimisation problem, where we need to assign a set of staff to a set of tasks given constraints such as skillsets, availability and working time capacity. These techniques can be used to allocate a workforce to tasks in the most efficient way.

In this notebook we will demonstrate K-Means Clustering using data stored in Clickhouse. We use the Online Retail dataset from the UC Irvine Machine Learning Repository. The dataset contains 541,909 transactions of a UK-based online store taking place between 01 December 2010 and 09 December 2011. The aim will be to cluster them into two or more groups based on spending patterns.

In this dashboard we analyse a customer dataset stored in ClickHouse Cloud to understand when and which classes of customer are churning. This dashboard was built to demonstrate the Metabase business intelligence tool.

In this dashboard we analyse dataset about ontime flight performance using the Looker Studio business intelligence tool.