Hello there!

I am a data scientist at a London-based tech startup. My main focus is in forecasting cargo movements and energy market behaviours around the world. This process involves a fair amount of time-series modelling, satellite image analysis and more general clustering problems.

I was previously a postdoc at the University of Illinois and the University of St Andrews where I also completed my PhD research in observational astrophysics. My main research goals are broadly concerned with understanding the formation, dynamics and kinematics of Active Galactic Nuclei (AGN). I have developed a Monte Carlo Markov Chain (MCMC) code to fit the time-series flux variations observed in AGN. This code, CREAM (Continuum REverberation AGN MCMC), uses bayesian inference to fit the Fourier time-series that drives the flux variations. It also infers the accretion disc and emission line time-lag response functions and returns the posterior probability distribution for all these parameters. The response function maps the accretion disc tilt-angle and radial temperature profile and allows us to probe the accretion disc structure for objects far too remote and compact for standard observations to be effective.

Please follow the following Links to mt CV, thesis, and publications. CREAM has recently been upgraded to a Python module (PyceCREAM) and can be installed from the terminal by typing pip install pycecream.

Data Projects



When not lost in space, I enjoy modelling a variety of data sets using various machine learning techniques including K-means clustering, multi-layer neural networks, random forrest classifiers and regressors and various other techniques from probability theory. The projects apply these techniques to a wide variety of datasets with applications anywhere from classifying underwater sonar signals as mines, to modelling variables affecting house prices in the US, to analysing shopping habits on Black Friday. The links below take you to my GitHub folder which contains the individual projects and codes (mainly python based). The 'projects' folder contains a short summary of the analysis process and main results of each project.

Active Galaxies: The Big Picture

Super massive black holes (SMBH) lie at the centers of almost all galaxies. These sometimes eat orbiting clouds of gas that wander too close and become Active Galactic Nuclei (AGN). The accreting material forms into a disc structure around the SMBH in a similar way to X-ray binary stellar mass black holes in our own galaxy. These discs are too remote and compact to be resolved directly and much about their structure and dynamics remains unknown.

1) Only an extremely massive and compact object (i.e. a black hole) can give rise to the observed high-speed and short pericenter (radii) orbits of the inner-most stars around the Milky Way centre (Schodel et al, 2002) . Even more exciting is that these SMBH's are present in most all galaxies in the Universe.

2) A small fraction of these black holes (<10%) consume material from their host galaxy and are called 'Active'. This material heats up as it falls into the black hole (sometimes outshining all the stars in the galaxy).

3) The region around (and including) the SMBH is known as the Active Galactic Nucleus (AGN). It is around 1 parsec (3.26 light years) in size and too remote to directly resolve using telescopes. AGN variability studies (Bentz et al, 2010 ; Shappee et al, 2014; Fausnaugh et al, 2016) and spectral observations (Mehdipour et al, 2015) allow astronomers to piece together cartoons (Urry and Padovani, 1995) that summarise our 'best guess' on the nature of these inner regions. See e.g. Padovani et al, 2017 or Netzer et al, 2015 for reviews.

5) I perform such variability studies (Starkey et al, 2016 ; Starkey et al, 2017) . I am interested in the accretion disc variability and study the flux variability (the light curve) across infra-red, optical, ultraviolet and X-ray wavelengths.

6) Astronomers observe that variations at shorter wavelengths tend to lead those at longer wavelengths Cackett et al, 2007; . Line variability (from the broad line clouds) lags continuum variability (from the disc) Peterson et al, 2004. This lets us constrain the size of the disc, BLR and black hole mass (De Rosa et al, 2013).



My research goals are broadly...

  • To understand the structure and dynamics of the accretion disc and surrounding broad-line-clouds.
  • Use this to help discover how the black hole is able to draw in huge amounts of material (around a solar mass each year) from its host galaxy.
  • What causes this to happen in some galaxies but not others?
  • Use AGN as cosmological distance measures (standard candles) to improve our knowledge of cosmology (See Watson et al, 2010; Hoenig et al, 2017).

CREAM (Continuum REverberation AGN MCMC)



PyceCREAM can installed using ''pip install pycecream'' and the source code can be found by following the link to the github repository below. Have fun :)

Useful & Interesting

Conferences & Meetings:

  • Outreach talk introducing black holes for general audiences - various venues
  • Back to the future: Astronomy with next generation instruments and simulations - Durham, UK - 8-9 Jan 2018
  • Quasars at All Cosmic Epochs - Padova, Italy - 2-7 Apr 2017
  • The Advantages of Resolution (13th Durham-Edinburgh eXtragalactic (DEX) Workshop with St. Andrews and Lancaster) - Edinburgh, UK - 9-10 Jan 2017
  • AGN STORM Meeting - Reykjavic, Iceland - 11-13 Jul 2016
  • AGN: What's in a Name? - Garching, Germany - 27 Jun - 1 Jul 2016
  • AGN STORM Meeting - Columbus OH, US - 14-16 Jul 2015
  • UK National Astronomy Meeting - LLandudno, UK - 5-9 Jul 2015
  • Quasar Day - Edinburgh, UK - 25-26 Jun 2015
  • American Astronomical Society 225th Meeting - Seattle WA, US - 4-8 Jan 2015
  • Cormak Astronomy Meeting - Edinburgh, UK - 27 Nov 2014
  • UK National Astronomy Meeting - Portsmouth, UK - 23-26 Jun 2014
  • Cormak Astronomy Meeting - Edinburgh, UK - 7 Nov 2013
  • Contact Details