Sevvandi Kandanaarachchi

Lecturer, Mathematical Sciences

RMIT University


Hello! Thanks for visiting my site! 👋

I am a mathematician working on data science related topics such as anomaly detection, meta-learning and streaming data. I conduct research on theoretical and real-world problems. For example, visualizing anomalies in high dimensions is a very tricky business. How can we reduce dimensions in a way that anomalies are highlighted or brought to the forefront? (For details, see dobin).

I also like working on real world problems, especially ones that are motivated by industry. From 2016 to 2019, I worked with an industry partner on intrusion detection. I’m also interested in bushfire mitigation research.

I came to statistical learning from a pure mathematics background. My PhD was in mean curvature flow, which lies in the intersection of partial differential equations and differential geometry. I like to bring my geometric intuition to solve data science problems.

More details can be found on my CV.


  • Data Science
  • Anomaly Detection
  • Streaming Data
  • Event detection
  • Meta-learning
  • Dimension Reduction


  • Graduate Certificate in Data Mining and Applications, 2015

    Stanford University

  • PhD in Mathematics, 2011

    Monash University

  • MSc Preliminary in Mathematics, 2007

    Monash University

  • BSc Eng. in Computer Science and Engineering, 2002

    University of Moratuwa




Mathematical Sciences, School of Science, RMIT University

Feb 2020 – Present Melbourne City

Research Fellow

School of Mathematics and Statistics, University of Melbourne

Jan 2020 – Feb 2020 Parkville, Melbourne

Research Fellow

Department of Econometrics and Business Statistics, Monash University

Jan 2018 – Dec 2019 Clayton, Melbourne

Research Fellow

School of Mathematical Sciences, Monash University

Jan 2016 – Dec 2017 Clayton, Melbourne

Assistant Professor, Mathematics

DigiPen Institute of Technology, Singapore

Aug 2011 – Aug 2015 Singapore


Dimension reduction for outlier detection using DOBIN

Instance Space Analysis for Unsupervised Outlier Detection

Recent & Upcoming Talks



Algorithmic IRT - an R package.


Event detection and early classification for streaming data - an R package.


Dimension reduction for outlier detection - an R package.

Recent Posts

Which nonlinear dimension reduction methods preserve outliers inside a sphere?

In this post we will look at how well nonlinear dimension reduction techniques preserve outliers that are placed inside a sphere, when …

Anomaly detection dilemmas

Finding anomalies/outliers in data is a task that is increasingly getting more attention mainly due to the variety of applications …

Tips for increasing your happiness during self-isolation

Teaching first year students made me realize how difficult it is for young people to stay at home all the time, specially during this …

Using dobin for time series data

The R package dobin can be used as a dimension reduction tool for outlier detection. So, if we have a dataset of \(N\) independent …


Gael Martin's podcast on Bayesian Statistics

During the 2020 working-from-home period Tim Macuga from ACEMS and I interviewed Prof Gael Martin from Monash University via zoom and …

Rob Hyndman's podcast on forecasting

Are you interested in forecasting? Well, Anthony Mays and myself did a podcast with Prof Rob Hyndman from Monash University on this …

Patricia Menéndez chats about Antarctica

Before the 2020 self-isolation period (also known as the lockdown) Tim Macuga from ACEMS and I informally interviewed Dr Patricia …

Space junk podcast

We humans leave such a lot of junk in space. If you’re interested in space junk, you can listen to two podcasts that I …