Winning Canada's Largest Hackathon
In September 2022, our team of high school students were a finalist at Hack the North, the largest Canadian university hackathon. We built ConnexSci, our solution to materializing research in the private domain.
There are fundamental barriers in the conventional research and scientific funding model that need to be addressed to accelerate the pace of discovery and commercializing research.
- Large funding bodies have made the process of getting funding bureaucratic over time and slow, and researchers are forced to spend the majority of their time applying for funding and design their experiment around a limited, fixed set of funds. This makes the process of funding more complex and it make sit harder for new and promising research to gain traction.
- The translation stage from early research to commercial pipelines (startups, biopharma) has become more complex, and terms + agreements around transfer strongly incentivize close-guarding IP from scientists. Terms are unfavourable to cause investors to lose interest in financing translation efforts.
- Venture Capitals and other LPs have little incentive and limited pool of possible companies and initiatives to finance because the change of success is hard to evaluate at such an early stage + not all technologies/research are able to return high returns. Big Pharma’s direct incentive to develop and acquire drugs with the largest possibilie market while miniziming R&D costs but this can skew research to areas that are established without assessing market need.
Our goal was to leverage Deep Learning and Network Insights to support new financial mechanisms for funding research that is important but incompatible in the status-quo funding space. The primary metrics are speed, ROI, and producitivity of the funds that are being used in said research.
We were deeply inspired by tools like Elicit, that summarize large volumes of research into readable information. We also found interest in interfaces like Obsidian, that can organize networks of knowledge into graphs.
Hackathon iteration of ConnexSci
Although this was only a hackathon project, we were quickly introduced to the problem of scale. We had to manage hundreds of thousands of research papers and build deep connections between them to repurpose existing knowledge. We approached this using distributed nodes:
We also chose to use use GraphML and NLP to identify high-impact research projects and allocate the necessary capital for producing effective therapies while reducing the financial and regulatory bottlenecks to translate early-stage drug discovery research into commercial products.
Electrical and Computer Engineering Faculty, University of Waterloo
The knowledge graph connects three types of nodes:
Category. Paper connections were illustrated in the following JSON:
Visualization with NeoVis
Interacting with nodes used their individual IDs to represent its data and edges.
The newest iteration of ConnexSci focuses on identifying core data within individual papers, instead of relying on a single global knowledge graph.
Latest version of ConnexSci
Identifying backlinks and relevant data help build neighbourhoods of knowledge around relevant data.