| Current Focus & Future Goals |
Machine Learning & AI
Like most science fiction fans, may view on artificial intelligence is that its both terrifying and utterly fascinating. Every time I read the tech news, I find something else to love about machine learning and AI, from language translating earbuds, to detecting the small changes in melanoma, to understanding how the face of the planet is changing and measuring crop yields, to convolutional neural network (CNN) generated abstract art. And of course there is astronomy, where MLA will undoubtably play a critical role in processing and analysing the seemingly unimaginable volumes of data — the dishes of the Square Kilometre Array (SKA) will produce 10 times the global internet traffic — expected from the next generation optical and radio telescopes. Over the past few years the industry has exploded, partly due to the availability of GPUs that make parallel processing ever faster, cheaper, and more powerful. It seems like everyone is trying to get a better handle of how machine learning and deep learning can solve the problems of tomorrow.
A short primer for the uninitiated:
Machine learning is all about using algorithms to parse data, learn from it, and then make a prediction. The algorithms enable computers to find complex relationships and patterns in data, and they produce outputs or decisions based on statistical models. Artificial neural networks (ANN) uses discrete layers, connections, and directions of data propagation to simulate the neural connections in the brain. Neural networks have been around for decades, but their potential was always computationally and algorithm limited. During my honours in year at the University of Melbourne, the Photonuclear Group tried to implement neural networks into their SNUPA parcel bomb detector. Training sets – signatures of the various components of explosives, or drugs – were small, the neural network had a single layers, and computationally the process was inefficient. Today it’s a different story. Deep learning uses neural networks with many layers, in some cases thousands of “hidden” layers, hence the term “deep”. The field is advancing so rapidly it’s hard to keep up. But almost every leading company, organisation, scientific body that has vested interest in machine learning has released a report or white paper outlining the current technologies and the promise of transformation for the better.
What am I most excited about?
Machine/Deep learning for Global Development and Humanitarian relief:
Monitoring troop build-up and movement of civilians in trouble spot and vulnerable regions; Using machine learning to model and potentially destabilise human trafficking; Predicting drought and crop yields in venerable areas; Monitoring coastal plastic pollution and areas where maintaining an ongoing supply of freshwater is critical; Potential for machine learning to provide more accurate (less false positives) anti-personnel landmine detection using GPR and other imaging datasets.
Machine/Deep learning in Science and Healthcare:
The ability to select representative samples of astronomical objects from the seemingly unimaginable volumes of astronomy data and to be able to detect and infer physical parameters and underlying physics; Potential improvements in medical imaging analysis and early detection, for example melanoma; Developing new products for preventative healthcare e.g., clinical decision support systems for quicker medical diagnosis, particularly important for developing countries where there may be limited access to medical facilities. A really interesting but controversial area is the ability to support mental health, e.g. Amelie.ai; Purely academic research producing useful products e.g., asbestos detection using ML with microscopy.
Natural Language processing:
Many of the above projects make use of NLP, but the technique itself if really exciting. It’s a challenging, problem, particularly when the context and nuances around language need to be understood. Recent exciting developments include language response and in-ear language translation devices.I’ve also seen a lot of really creative projects based on literature and linguistics research.
Deep Learning for artistic & creative output
CNN trained on photographic images, stores visual information in some layers as well as other more abstract features, such as edges, shapes and gradients. The visual output being abstract representations of the trained images and classes. There is a sub-field in AI called computational creativity that questions whether a machine can be creative, or exhibit creative behaviour. Aside from the aesthetic there may be good philosophical reasons for exploring this
- Creating machine learning projects and tutorials using Python’s Natural Language Toolkit (NLTK)
- Creating astronomy machine learning projects and tutorials using the AstroML Python library and SDSS and HST data.
- Creating deep learning projects and tutorials using Tensor Flow
Free to Feed
| Tech Solutions for Social Innovation |
Free to is a Melbourne–based social enterprise run by Loretta and Daniel Bolotin, that recognises the entrepreneurial characteristics and existing skills of refugees and new migrants, as well as the significant challenges that they face in gaining meaningful employment, or pursuing new enterprises in Australia. The heart of Free to is its pop-up cooking school Free to Feed, where all classes are run by highly skilled refugees and asylum seekers.
My first experience with Free to Feed was in early February 2017 when my mum and I took Hamed's Persian Vegetarian cooking class. I thought this would be the perfect Christmas present for someone who has it all, and it was. While it was immediately obvious that Free to were experts in this sector and onto a really good thing, I couldn’t help but notice some issues around business processes and the technology they were using. Their booking system was clunky; choosing a class based on location, date, chef or cuisine was difficult, and in our case it required a few emails back and forth, and some manual intervention to successfully input my gift certificate. Booking by location wasn’t an option and since my mum lives an hour away from the city this something we had to consider. Since classes are held all over Melbourne it seemed like a really obvious and potentially simple thing to have.
I knew then that Free to Feed could really benefit from the Random Hacks of Kindness (RHoK) treatment. Talking with Daniel, it quickly became apparent that their booking system and website were only a small part of the problem. One of the biggest roadblocks to growing the business was their inability to scale, and scale quickly. They had received quite a lot of press coverage, after all Melbourne is a foodie city, and there was no shortage of customers. Communications were also an issue. Many of their processes are manual and require one-on-interactions with a diverse range of stakeholders and clients.
In May 2017 Free to Feed began working closely with RHoK. My role — alongside fellow RHoK organiser Tim Elliot (a strategic designer with a background in industrial design), was to guide Loretta and Daniel through the RHoK lifecycle and prepare them for each stage of the process. Unlike most hackathons, RHoK engages with change makers roughly 6-weeks prior to the hack weekend, and where possible supports ongoing development (RHoLLs) until change makers have a fully tested and deployed tech solution. In some cases engagement with RHoK lasts up to a year from initial contact.
As a RHoK buddy/Tech Lead, I help guide Free to through Agile software development and Cynefin Framework decision making processes, assist in the development of problem statements, introducing Free to user–centered design and facilitating Contextual Inquiry (CI) sessions with User Experience (UX) experts, and liaising with technologists, facilitating team development on hack weekend and ensuring they are well resourced to deliver a solid tech solution, and facilitating ongoing RHoLLs where appropriate.
| Challenging the Status Quo |
techsavvyastronomer.io is about changing the way we approach modern day, data–intensive astronomy research. It’s about challenging the status quo and creating a new culture of tech savvy astronomers, equipped with industry standard tech skills to complement scientific computing and domain specific data analysis expertise. It's also about preparing astronomers — and researchers from similar data–intensive research disciplines — for successful careers outside of academia.
The website is based around a directory of useful tools that researchers can use as part of their every day research, or to develop innovative tools for the community, or small web-based projects that showcase their research. Where possible I've included links to useful tutorials and online help. Tools are loosely grouped into categories – General Tools, Web Design, Data Visualisation tools, Website Hosting, Astronomy Tools – based on their core function. This side project was inspired by the successful .Astronomy7 – Day Zero research and tech skills training day that I organised. and the Web Development & Research Tools for Astronomers (PDF, 2MB) guide that I created for the conference.
Future Assembly 2016
| Big Data & Tech Talks |
At this year's Future Assembly technology festival, I spoke about data-driven discovery and the most ambitious telescopes ever built. The idea was to kickstart a discussion about the next-generation astronomical facilities and telescopes, how they are driving new technologies, and how they are changing the way astronomers approach data–intensive research. I also wanted to highlight some of the fantastic projects being led by Australian researchers, starting with the Square Kilometre Array (SKA). Anticipating a diverse audience, I kept this talk quite general, and created this webpage page for those who wanted more information. It's essentially a curated selection of videos from the SKA, LSST, GMT, JWST, and other NASA projects. I've also included links to the official project websites.
Astronomy Australia Ltd
| Advisory Work |
I serve on two of Astronomy Australia Limited's (AAL) committees, the Astronomy eResearch Advisory Committee (AeRAC) and Computing Infrastructure Planning Working Group (CIPWG). AeRAC monitors and assesses the performance of key HPC facilities, identify gaps in astronomy's eResearch access, and keep AAL apprised of any emerging opportunities. CIPWG was established by AAL in October 2015, and tasked with advising AAL on appropriate investments over the 5-year period (beginning July 2016), in order to achieve the computational and software goals in the Australian Astronomy Decadal Plan 2016-2025. I also contributed to the e–Science Working Group report for the Decadal Plan. As part of the CIPWG, I was responsible for drafting key recommendations around scientific computing training and support; specifically, recommending strategies (and providing an estimate of the level of investment required) to provide support and training to the astronomy community that maximised current resources and leveraged community initiatives. The final report was published in October 2016 as part of a call for tender, to deliver a new Astronomy Data and Computing Service (ADACS) that will provide e–Research support services for the astronomy research community.
ARC CoE for Gravitational Wave Discovery
| Freelance Consulting |
Australian Research Council (ARC) Centres of Excellence are prestigious research centres that undertake highly innovative and world-leading research of high international standing and impact. In September 2016, the ARC Centre of Excellence for Gravitational Wave Discovery (OzGrav) was awarded $31.3 million. Over the next 7 years it will capitalise on the first detections of gravitational waves to understand the extreme physics of black holes and warped space-time.
Conferences, Seminar Talks & Workshops
During the past 10 years I've given numerous invited seminar & contributed conference talks about my astronomy research. I've also given talks about best practice in research data management, scientific computing, how to get the most out of research hack days, and on the benefits of building a tech savvy research community.
At Swinburne I presented a number of high-level research strategy and research infrastructure talks, at faculty committee meeting, research technology forums and for Swinburne Research and the Deputy Vice Chancellors's research planning days. I have also hosted numerous workshops focussed on research tech skills, scientific computing, and creative coding.
Most recently I was a speaker at the Future Assembly technology festival in Melbourne. I talked about data–driven discovery and the technological advances and challenges associated with the next–generation astronomy facilities. Selected talks are available on Speaker Deck.
Professional Associations & Activities
I serve on a number of working groups & advisory committees, mainly focussing on big–data management, and requirements around scientific computing and research infrastructure. I was recently invited to join, as a founding member, a new IAU Working Group on Data–driven Astronomy Education and Public Outreach (DAEPO).
Throughout my career I've participated in, chaired, and established working groups and forums to develop innovative strategies and initiate change. I don't believe in joining advisory and working groups unless I can make significant contributions. I actively champion those that have a diverse membership. I've always found these to have the most interesting conversations, and as a result these tend to be the most productive. I've also sat on selection committees for academic appointments.
Scientific Papers, Proposals & Reports
I've co-authored numerous scientific research papers in high-impact journals. You can find my refereed publications and conference proceedings on the SAO/NASA Astrophysics Data System (ADS), the arXiv e-prints server, or on Google Scholar.
I've also written consultation and discussion papers, prepared responses to consultation papers and government reports, drafted terms of reference, working group reports, research communication summaries, policies and policy guidelines, technical reports, and documentation for software and research infrastructure services.
I wrote Swinburne University of Technology's policy and accompanying guidelines for the Management of Research Data and co-authored the Our Swinburne Institutes Model: Proposal, with the Deputy Vice Chancellor (Research & Development). The proposal is a detailed overview of the new Swinburne Institutes Model which came into effect mid–2016 with the establishment of five brand new research institutes – Data Science, Heath Innovation, Smart Cities, Social Innovation, and Manufacturing Futures – to foster interdisciplinary collaboration.
Creative Coding, Data Visualisation
& Machine Learning
| Independent Projects |
In August 2016, I attended a week-long summer school hosted by the Berkeley Institute for Data Science (BIDS) & GitHub. The focus was on effective computing, Bayesian statistics, machine-learning algorithms, and optimisation and sampling. Although the concepts were presented in an astronomy context using, in most cases astronomical data, the methodologies used were general enough for almost any scientific dataset. As part of the workshop I began a collaborative project to create a series of Simple Database tutorials for Pythonic Astronomers for graduate level researchers. I usually keep all my non–astronomy coding projects on GitHub.
I've started creating more general python (scikit-learn, matplotlib, numpy etc.) based tutorials, based on other projects that were pitched during the week. The first of these is a tutorial for Creating Color Palettes from Apollo Project (or any RGB) Images.
When Astronomy Meets Tech
| Colliding Worlds |
.Astronomy (dotastro) was created in 2008 by Robert Simpson, along with Alasdair Allan, Sarah Kendrew, Chris Lintott, Stuart Lowe, Carolina Odman Govender, Arfon Smith and others. I like to think that it started out as a small group of renegade astronomers who lamented the lack of blue sky thinking and did something about it. Based on my experience this seems like a fairly good description, particularly since of number of .Astronomy founders were instrumental to building the Zooniverse (a unique model where research is powered by citizen scientist's). Over the years .Astronomy has grown into a tight community of astronomers who have a passion for creative coding. Conferences are deliberately kept small, roughly 50 participants are selected based on application. Astro Hack Week has a similar model, which has also proven to work well. Spin-off astronomy hack days are now a regular feature of the National Astronomy Meetings in the UK (NAMhack) and the annual American Astronomical Society conference (check out#AAS225 & #hackass).
I organised the 2015 .Astronomy7 conference in collaboration with Amanda Bauer (@astropixie), Vanessa Moss (@cosmicpudding), James Allen (@j_t_allen), and Rob Hollow (@roberthollow). I also put together the pre-conference Day Zero research training day. The idea was to We aim to introduce a range of skills that will not only be useful for the Hack Day, but for astronomy research in general. I also created the first .Astronomy Day Zero Guide: Web Development & Research Tools for Astronomers. This has become a standard resource and modified for subsequent conference. It was also the motivation for creating the techsavvyastronomer.io website
Since I first attended .Astronomy6 back in 2014, I've slowly been building up my web development skills, starting with simple projects and then challenging myself to understand all aspects of development (back–end, front-end, design, UX/UI). I really enjoy visualising data in creative ways and telling stories through data. Projects that have been written up as blog posts can been found here, or on GitHub.
This year, I'm thrilled to be attending .Astronomy9 as an invited speaker. I will be talking about; Web Technologies for Data-Driven Astronomy & Big-Data Science.
One of the most challenging questions in observational astronomy is [still...] , how do clusters of galaxies evolve? How significant is the cluster environment, and what transformational and star formation processes govern the evolution of individual galaxies. For decades astronomers have been studying this problem and while we have certainly come a long way there is still much to understand. My expertise is working with high-resolution imaging and spectral data and developing data analysis routines and pipelines for large datasets.
In 2009 I moved to Liverpool and joined the HST/ACS Coma Cluster Treasury Survey. In November 2011, I moved to the University of Oxford to work with Roger Davies and Davor Krajnovic on a follow-up observations for the Atlas3D Galaxy Project. The goal was to model the three dimensional structure of galaxies, using high-resolution archival images from the Hubble Space Telescope (HST), Prior to this work, the project team had focussed on understanding the kinematic properties, relying only on the molecular and gas distribution of galaxies. That data was obtained using the SAURON optical integral-field spectrograph on the William Herschel Telescope (WHT). Our paper; ATLAS3D Project – XXIII: Angular momentum and nuclear surface brightness profiles, was published in 2013
e-Research at Swinburne
| Strategy & Building Capability |
My e-Research Consultant role at Swinburne Research (Feb 2013 – Feb 2016) was quite varied. I consulted on a wide range of e-Research infrastructure projects with Research Information Services and Swinburne ITS, developed data management policies and procedures, set up systems that promote and enable best practise in research data management, initiated grassroots initiatives that promote industry skills for researchers and alternative career paths, and setup software and development support for researchers.
As part of the Swinburne/ANDS Metadata Stores Project I developed the University's policies and guidelines concerning research integrity and data management, and meeting compliance requirements set by the Australian Research Council (ARC) and National Heath and Medical Research Council (NHMRC).
Typically, e-Research roles vary across Australian universities. The majority of my work was done in collaboration with the PVC (Research), PVC (digital Frontiers), DVC (Research & Development), and the Research Information Services team. By virtue of my research background, I've also maintained strong ties with Swinburne's Centre for Astrophysics and Supercomputing (CAS). My focus has always been on identifying and finding solutions to the current (and potential) issues that negatively impact data-intensive — "big data” — research.
Swinburne/ANDS Metadata Stores Project
In February 2013 I joined Swinburne as a Research Data Analyst/Librarian as part of the Swinburne ANDS Metadata Stores Project. Swinburne had successfully obtained funding from the Australian National Data Service (ANDS) for the development of MyTardis and ReDBox research data management systems, that would facilitate the storage and open access to select astronomy, atom optics and neuroscience datasets. I was responsible for overseeing the development and implementation of two MyTardis systems (for brain imaging and atom optics) and Swin ReDBox research metadata store. I also worked on the customisation and testing of the ReDBox system and the integration with Research Data Australia and the National Library Archive.
The overall goal of the initiative was to implement a number of systems with the potential to support university-wide solutions for the discovery, sharing and re-use of rich data collections. Swinburne's research data collections would also be fed directly to Research Data Australia, the primary data discovery service of ANDS.
Swinburne Chapter – The Hacker Within (THW)
The Hacker Within began as a student organisation at the University of Wisconsin-Madison, and is now reborn as a collection of such chapters around the world. Active chapters include Wisconsin, Berkeley, Yale, and Melbourne (Swinburne University of Technology). Each of the chapters convenes a community of researchers, at all levels of their education and training, to share their knowledge and best practices in scientific computing to accomplish their work.
Following a working visit to Berkeley Institute of Data Science, I launched Swinburne Hacker Within (SHW) in April 2015, as a weekly, multi-disciplinary, digital technology meet-up. The goal was to establish a forum for PhD students, postdoctoral researchers & academic staff, and research & data librarians, to discuss issues around "big-data" research, collaborative coding, the value of open–data, copyright and ownership of data, and reproducible science. It was also forum for sharing & discovering new tools, leading or participating in tutorials, and creating hack projects with researchers from other disciplines. Topics included IPython & Jupyter Notebooks, Julia programming, D3js data-visualisation, version control with git & GitHub, interactive plotting, mapping with Carto, crisis mapping for disaster relief, creating websites for GitHub projects, SQL, and relational databases.
Two reasons why The Hacker Within initiative is important:
- To prepare researchers for alternative careers in the technology industry. The rise of fellowship programs, for example Insight Data Fellows and Science to Data Science, enable scientists to learn the industry specific skills needed to work in the growing field of big data at leading companies. With new skills in data science and software development, scientists with analytical backgrounds are now in great demand on the European and US job market and are being offered jobs in leading tech-companies.
- The tenets of scientific research (e.g. data control, reproducibility, and peer review) suffer in projects that fail to make use of current development tools such as version control, testing, and comprehensive/automatic documentation. To avoid these pitfalls, the numerous Hacker Within Chapters exist for the purpose of sharing skills and best practices for computational scientific applications.
University Institutes Model
During my last 6 months at Swinburne, I worked closely with the Deputy Vice-Chancellor (Research & Development) as project manager to implement the new Swinburne Institutes Model. This bold and exciting plan will help Swinburne to reach its full potential, by aligning its efforts more closely with the emerging national research and innovation priorities, growth industry sectors, and strategic research engagement opportunities in industry and business.
The Swinburne research institutes institutes will work at the frontiers of research and innovation, with multidisciplinary teams tackling big challenges with potential for transformative economic and social impact. They cover a diverse range of disciplines – Data Science, Health Innovation, Smart Cities, Social Innovation & Manufacturing Futures. Each institute is set up to foster interdisciplinary collaboration within faculties, research centres, researchers in industry, business and community. Their approach is outward oriented, and outcomes and impact focused.
Random Hacks of Kindness
| Tech Solutions for Social Impact |
Random Hacks of Kindness – Australia, is a part of a global community of technologists and change makers hacking for humanity. The Random Hacks of Kindness (RHoK) movement began in 2009 as a joint initiative between three tech companies – Microsoft, Google, and Yahoo! – NASA's Open Government team and the World Bank's Disaster Risk Management Unit. The idea was to organise a hackathon where developers and data analysts could produce open source, technology solutions, useful in the event of (or to mitigate the impact of) major disasters around the world. The first event was held at the Hacker Dojo in Mountain View, California, and christened "Random Hacks of Kindness".
Over the years, simultaneous satellite events have established local RHoK communities around the world. These tend to focus more on developing solutions locally rather than globally, working closely with grassroots organisations, non-profits, community groups and socially conscious individuals keen to make a difference. Australia has one of the most vibrant and engaged RHoK communities in the world with RHoK Chapters in Melbourne, Sydney, Brisbane, and in the regional centres of Ipswich, Western Sydney, and most recently Bendigo. In fact, RHoK Australia is the largest and longest running hackathon for social good in Australia. I've been on the Melbourne RHoK Organising Committee for the past two years.
Each year the hackathons each bring together roughly 70 people from a variety of backgrounds who want to contribute in some small way. Business analysts, software developers, web and application developers, designers, scientists and educators, writers and communicators, projects managers and UI/UX testers. Generally clever, interesting, and motivated people with fresh ideas. When I first began working with RHoK, events were directed by Angus Hervey (@angushervey), who has since moved on to co-found Future Crunch. The current the National Community Manager for RHoK is Cal Foulner (@CalFoulner), co-founder of Beanstalk and Forage. RHoK is supported by a dedicated National Steering Committee, and city-based Organising Committees. In mid-2015, I joined the Organising Committee for RHoK Melbourne. Helping shape this community is a real privilege, and it's been great to see RHoK Australia's portfolio of projects grow over the past year. To date Local Linguist, Changing Places (i.pee.freely), Right Click Community, and Free to Feed are my favourites.