Dylan's Photo

Dylan Sandfelder

About

I am a PhD student at the University of Oxford doing a DPhil in Engineering Science. I was recently a resident applied ML engineer at Kumo. In 2022, I graduated from the Oxford with an MSc in Advanced Computer Science where my thesis for that program received a Distinction. Before that I got a BSc from McGill University doing Honours Computer Science and a Minor in Mathematics. I graduated from that program in 2020 with Distinction.

My DPhil at Oxford is supervised by Prof. Xiaowen Dong in the Department of Engineering Science and Prof. Mihai Cucuringu in the Department of Statistics.

As a master's student at Oxford I worked with Dr. Ismail Ilkan Ceylan to complete my master's thesis. The thesis explored a novel method of tackling inductive learning tasks on knowledge graphs using graph neural networks and knowledge graph embedding models. The thesis received a Distinction.

As an undergraduate student at McGill I did research at both McGill and Mila in the field of graph neural networks with Prof. William Hamilton after receiving a McGill Science Undergraduate Research Award. The work explores the use of ego-nets to improve the capabilities and power of GNNs. We submitted a paper (of which I am the first author) on this work that was accepted to a special session of IEEE-ICASSP 2021. That paper can be viewed here. The work I did as Mila was also followed by a workshop paper for the NeurIPS 2021 Data-Centric AI workshop. The paper explores the taxonomy of graph datasets and can be viewed here. My profile with Mila can be found here.

I was also the CTO and co-founder of Piriko, a real-time event planning app that allows users to follow and engage with events happening live in the present. Our amazing 6 person team launched versions of the app on the iOS Store and the Google Play Store. We entered into a agreements with Concordia University to work with them on establishing thriving student communities. The school published an article about the startup that can be viewed here.

I am a co-founder and the CTO of Amorphous AI. Clinical data is unstructured. We are building solutions that automate healthcare data structuring and analysis. Doctors should focus on the patients, we take care of the rest. We believe that Europe is a great place to build, and we are lucky to have made a team of AI nerds who won't fall asleep until the benchmark is nailed.

I have a lot of international professional experience after having been raised in Shanghai and Singapore in a bi-lingual home (English and French) before doing my undergraduate degree in Montreal.

Experience

Startups
2024 - Present: Amorphous AI (London, UK) - Co-Founder, CTO
2022 - 2023: Kumo (Mountain View, CA / Oxford, UK) - Resident Applied ML Engineer
2018 - 2022: Piriko (Montreal, QC) - Co-Founder, CTO
Internships
2025: G-Research (Spring into Quant Finance) (Nice, France) - Quant Intern
2024: Record Financial Group (London, UK) - Quantitative Analyst Intern
2020 - 2021: Mila, i.e. Montreal Institute of Learning Algorithms (Montreal, QC) - Research Intern
2018: Medtronic, Surgical Robotics Lab (Boston, MA) - Software Engineering Intern
2017: Kouzhu Educational Technologies (Nanjing, China) - Lecturer, Educator
Organizations
2016 - 2017: McGill Robotics (Montreal, QC) - Section Leader, Mentor, 2017 International RoboSub Finalist
Distinctions
Oxford-Man Institute of Quantitative Finance - Full DPhil Funding
Oxford MSc Thesis - Distinction
McGill BSc - Distinction
McGill Science Undergraduate Research Award
International RoboSub Competition Finalist

Publications

Sandfelder, Dylan and Cucuringu, Mihai and Dong, Xiaowen, "Data-Driven Graph Filters via Adaptive Spectral Shaping," GSP 2025 - Graph Signal Processing Workshop (GSP), 2025
Oliveira, Daniel and Sandfelder, Dylan and Fujita, André and Dong, Xiaowen and Cucuringu, Mihai, Tactical Asset Allocation with Macroeconomic Regime Detection (March 18, 2025). Available at SSRN: https://ssrn.com/abstract=5183762 or http://dx.doi.org/10.2139/ssrn.5183762
D. Sandfelder, P. Vijayan and W. L. Hamilton, "Ego-GNNs: Exploiting Ego Structures in Graph Neural Networks," ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2021, pp. 8523-8527.
Liu, Renming, et al. "Towards a Taxonomy of Graph Learning Datasets." Presented at the Data-Centric AI Workshop at NeurIPS 2021, December 2021.

Something fun... Conway's Game of Life

Go and look for a new browser.
Current Score: 0
Curriculum Vitae