Led the research, design and development of a novel analytical process and product suite for reliably and quickly categorizing mineral assemblages in a set of rock core samples using a hyperspectral imaging laboratory for use in very remote and mostly offline sites.
Hyperspectral geology makes use of light-detecting instruments to identify minerals in rocks. This is (somewhat) analogous to how our eyes capture light and our brains analyze the signal to identify that a banana is ripe (yellow) and ready to eat as opposed to when it is under-ripe (green).
Developed a suite of products for mapping minerals in geological core samples using hyperspectral-imaging technologies. Managed/supported R&D team, grants and intellectual property portfolio. Led and actively participated in the development of numerous completed products integrating hardware and software. Co-developed new directions for the business strategy. Actively shaped and evolved the company culture.
The majority of geological mapping of rock core samples is done without any on-site analytical support, and often using spreadsheets. Given that the labour costs and value of results are extremely high to companies engaged in mapping of core, the opportunity to improve the reliability of the mapping and the speed of the process is tremendous. The value proposition was that we could provide reliable mapping and highly portable data for about a tenth of the cost of the geologist per meter of core and also in a quarter of the time. Note that the goal was not to replace the geologist, but to greatly facilitate their work by improving speed, reliability, ease-of-use, and ease of data sharing.
Approximate time to analyze 500m of core rock samples
By the on-site hyperspectral imaging lab
By an on-site trained geologist
By an off-site laboratory
I joined the company as the third employee, working initially as a signal specialist. As I grew into the role of Director of Research and Development, and Production Capacity, I took on a wide variety of roles but continued to design, develop, and prototype the scientific software and processes responsible for analyzing the data and producing a final product.
Here are many of the specific tasks which I was responsible for throughout the process:
- Product & UX design from beginning to end (from a box of rocks to scientific results)
- Scientific research
- Grant applications
- Designing, making and testing opto-mechanical prototypes
- Researching and prioritizing new technologies to develop
- Team leadership
- Purchasing and ordering
- Product development
- Project and project portfolio management
- Technical software support for remote teams
- Software development for hyperspectral geoscience
- Evaluating algorithms for scientific and production efficiency
- Python libraries used: numpy, matplotlib, scipy, h5py, opencv
- Domains: image & signal processing, data processing & visualization
- Data visualization and scientific UX/UI design
The imaging device was set up in a trailer as a mobile imaging laboratory. I worked closely with the parent company that designed and produced the imaging instruments (Photon Etc), and a variety of staff in-house, including marketing and communications, geologists, lab technicians, and on my team, engineers and software developers.
We used Trello to record bugs and feature requests – we tried to make it easy to understand and stay very responsive and receptive to input from all the other departments and staff functions. Our lab technicians and geologists provided us with the majority of live testing and invaluable feedback. Providing them with a clear roadmap of improvements and extensive training was essential to both a resilient work culture in the face of unexpected challenges and to the product success.
Images of the developed software and UI sketches
Title: System for analyzing and categorizing a geological drill core sample. CA2813913A1. Date: 2012-08-21. Patent status: Lapsed. Description: Use of a hyperspectral imaging device and supporting software to assist geologists in analyzing and categorizing geological drill core samples.
Planning out and prototyping major product improvements
For the first version of the instrument, the data which came from the imaging system required numerous software post-processing steps. Many of these were due to the complication of capturing both (x,y) spatial information and spectral information at the same time. The relatively long scanning times were compensated for by placing a larger number of samples in the field of view. However, the unintended consequences were that the lab had to be considerably larger, require more electricity (for all the lights), and cost more to produce the labs.
A second version was designed and prototyped using a line-scanner and motorized conveyor (shown below). There was not sufficient time to fully develop the new system before Photonic Knowledge ran out of cash, however, the early results of the system in terms of analytical accuracy and precision, data production speed, low complexity, and system cost were extremely promising.
- Whereas we began with an instrument and a rudimentary instrument control interface, over the course of 5 years, I successfully led a team through the design and development of a cohesive product suite, analytical process and client-facing software. During this time, I developed various innovations, including one filed patent regarding the easy-to-use presentation of analytical results, another prepared (but unfiled) patent regarding the data calibration and correction process, and a third innovation to the algorithmic identification process.
- Developing the software suite for in-house use by lab technicians gave us so much flexibility to cope with and resolve unexpected problems. Regularly listening to, gathering feedback from, and communicating progress to the lab technicians and geologists was essential to the development process.
- Photonic Knowledge received the Deloitte Technology Fast 50 Award in 2012.
- The vitality of the mining industry strongly influenced the company’s evolution. It began on the heels of the global financial crisis of 2008-2009, followed by commodity prices increased to a peak in 2011 and a subsequent decline by 2015 back to 2008-2009 lows. With a shortage of new contracts in 2013 and continuing into 2014, there was insufficient cash flow to float through the dip—ultimately, the company closed in 2014.