Fynn Schmitt-Ulms
Machine Learning Engineer
Hi, I am a machine learning engineer at Themis AI, in Cambridge, Massachusetts. I am developing Capsa-Torch, a python package that automatically transforms machine learning models into "risk-aware" variants. I am interested in making machine learning more robust and safe so that it can be applied to real world problems.
Prior to joining Themis AI, I received my B.Sc. with First Class Honours at McGill University.
Experience
Machine Learning Engineer | Themis AI (2023-Present)
Lead Developer of Capsa-Torch: a python package for the automatic transformation of PyTorch Modules to “risk-aware” variants through the application of uncertainty-estimation research
Gained expert understanding of PyTorch internals and SOTA uncertainty-estimation approaches
Designed and implemented transformation algorithms in a way that is robust to the vast variety of different neural network architectures and implementations
Tested solutions on test set of ~14k most popular PyTorch Modules on GitHub
Competitive Programmer (unpaid) | McGill University (2021-2022)
Member of McGill’s Official Competitive Programming Team
Placed 11th in Northeastern North America Region of ICPC
Helped develop training program for future students
Student Researcher | Bielefeld University (Summer 2021)
Funded by Mitacs through the DAAD Rise Scholarship Program
Developed the winning solution for Track 2 of the AI for TSP Competition
Presented work at IJCAI: Data Science for Optimization Workshop
Teacher Assistant | McGill University (Winter/Fall 2021)
Third-Year Computer Science Course: Programming Languages and Paradigms
Hosted weekly online tutorials and office hours for students
Graded quizzes and assignments and monitored discussion boards
Student Researcher | University Health Network (May - August 2019)
Worked in a bioinformatics research lab (35 hours each week)
Used R to analyze large pharmacogenomics datasets and generate insights
Worked with server clusters and created parallelized/GPU-runnable code for faster computation
Publications
Yang, Q.; Ravikumar, S.; Schmitt-Ulms, F.; Lolla, S.; Demir, E.; Elistratov, I.; Lavaee, A.; Lolla, S.; Ahmadi, E.; Rus, D.; Amini, A.; Perez, A. (2023) Uncertainty-Aware Language Modeling for Selective Question Answering. arXiv:2311.1545. http://arxiv.org/abs/2311.15451.
Perez, A.; Elistratov, I.; Schmitt-Ulms, F.; Lolla, S.; Ahmadi, E.; Rus, D.; Amini, A. (2023) Risk-Aware Image Generation by Estimating and Propagating Uncertainty. ICML: Challenges in Deployable Generative AI Workshop. https://openreview.net/forum?id=zzboa1TtNI.
Schmitt-Ulms, F.; Hottung, A.; Sellmann, M.; Tierney, K (2022). Learning to Solve a Stochastic Orienteering Problem with Time Windows. Learning and Intelligent Optimization; Lecture Notes in Computer Science; Springer International Publishing: Cham; Vol. 13621, pp 108–122. https://doi.org/10.1007/978-3-031-24866-5_8.
Zhang, Y.; Bliek, L.; Costa, P. da; Afshar, R. R.; Reijnen, R.; Catshoek, T.; Vos, D.; Verwer, S.; Schmitt-Ulms, F.; Hottung, A.; Shah, T.; Sellmann, M.; Tierney, K.; Perreault-Lafleur, C.; Leboeuf, C.; Bobbio, F.; Pepin, J.; Silva, W. A.; Gama, R.; Fernandes, H. L.; Zaefferer, M.; López-Ibáñez, M.; Irurozki, E. (2023) The First AI4TSP Competition: Learning to Solve Stochastic Routing Problems. Artificial Intelligence, Vol. 319, pp 103918. https://doi.org/10.1016/j.artint.2023.103918.
Experience
Honours Computer Science B.Sc. | McGill University (2018-2022)
4.0 cGPA
Dr. Feng Qian Convocation Prize for excellence in Computer Science
Hugh Brock Major Entrance Scholarship
High School | Senator O'Connor College School (2014-2018)
TCDSB Top Scholar: Top Grade 12 Average in School Board
Governor General Award: Top Grade 11 & 12 Average
Proficiency Award: Top Grade 11 Average
Coding
Please see personal projects for some examples of my work.
Github: https://github.com/fynnsu