CV
Last updated: Jul 2024.
[Download full version]
Education
ETH Zürich
MSc ETH in Computer Science, Sep 2020 – Mar 2023
- Major in Machine Intelligence and minor in Programming Languages and Software Engineering.
- Thesis: “Exploring Data Collection Dynamics Through Data Valuation.”
University of Copenhagen
BSc in Computer Science, Sep 2017 – Jun 2020
- Specialization in Data Science with coursework focused on machine learning.
- Thesis: “The Carbon Footprint of Training Deep Learning Models.”
University of California, Merced
UCEAP Reciprocal Program, Aug 2019 – Dec 2019
- Exchange study part of my BSc CS at UCPH with coursework focused on machine learning.
Experience
UBS
Quant Developer, Jun 2023 – Present
- Design and develop big data tools and solutions for Treasury Risk Control’s balance sheet analytics.
- Lead developer for a library calculating cash flows from position-level data, enhancing risk management through detailed sensitivity analysis using automatic differentiation.
- Drive code infrastructure improvements, including CI/CD pipelines and migration to Databricks and Spark, enhancing data processing speed and reliability.
- Implement machine learning models for predictive analytics and risk assessment, resulting in more accurate forecasting and better-informed risk management decisions.
Alexandra Institute
AI / Machine Learning Specialist, Mar 2023 – Jun 2023
- Dual role in applied research and expert consultancy in machine learning, focusing on natural language processing and utilizing pretrained transformers.
University of Copenhagen
Teaching Assistant, Jan 2020 – Jul 2020
- Assisted in teaching the Data Science course, covering databases, machine learning, and data pipelines.
Nykredit
Software Developer, Oct 2018 – Jan 2020
- Built financial software for internal advisors in an agile C# development team.
- Developed and maintained financial software for mortgage loans in .NET, reducing processing time.
- Implemented continuous deployment pipelines with Jenkins and BitBucket, fully automating integration testing and deployment, which improved deployment efficiency and reliability.
Publications
Carbontracker: Tracking and predicting the carbon footprint of training deep learning models
L. F. W. Anthony, B. Kanding, and R. Selvan
ICML Workshop on Challenges in Deploying and monitoring Machine Learning Systems, 2020
Paper CodeCarbon footprint of selecting and training deep learning models for medical image analysis
R. Selvan, N. Bhagwat, L. F. W. Anthony, B. Kanding, and E. B. Dam
International Conference on Medical Image Computing and Computer-Assisted Intervention — MICCAI 2022, 2022
Paper
Highlighted projects
Carbontracker
github.com/lfwa/carbontracker
Open-source tool for tracking and predicting the energy consumption and carbon emissions of training deep learning models in Python. The tool is freely distributed under the MIT License. Corresponding publication arXiv. It has been downloaded $>$$75$k times on the Python Package Index (PyPI) as of writing.
Datadynamics
github.com/lfwa/datadynamics
Open-source library and environment for simulating data collection dynamics in multi-agent settings, primarily targeting the exploration of data valuation approaches. The library is freely distributed under the BSD 3-Clause License.
Reinforced Graph Neural Networks for Collaborative Filtering
github.com/lfwa/reinforced-gnn
Introduced a novel architecture to generate predictive compatibility scores for never-before-seen content in recommendation systems. The architecture combines the strength of graph-extracted embeddings in a graph neural network with the generalization power of a deep feed-forward network and adds “reinforcements” providing additional information to the network.
Static Taint Analysis For Ethereum Contracts
github.com/lfwa/vulnerable-ethereum-contracts
Designed and implemented a static taint analyzer in Datalog for Ethereum smart contracts. The analyzer detects vulnerable contracts that may be deleted from the blockchain and have all remaining cryptocurrency transferred to an untrusted address.
Supporting Alternative SMT Solvers in Viper
github.com/viperproject
Added support for multiple SMT solvers, such as cvc5, in the symbolic-execution based automated verification backend written in Scala for the program verification tool chain and infrastructure, Viper.
Relevant coursework
Machine Learning & Big Data | Mathematics | Software Engineering |
Advanced Machine Learning | Statistics & Probability Theory | Program Verification |
Causal Representation Learning | Discrete Mathematics | Program Analysis for System Security and Reliability |
Natural Language Processing | Linear Algebra | Concepts of Object-Oriented Programming |
Probabilistic AI | Modelling & Analysis of Data | Computer & Network Security |
Reliable & Trustworthy AI | Algorithms & Data Structures | |
Computational Intelligence | Randomized Algorithms | |
Big Data |
Skills
Programming Languages
- Python, C#, SQL, Rust, Scala, F#, Java, C, Datalog
Databases
- PostgreSQL, Oracle
Frameworks and Tools
- PyTorch, TensorFlow, Gym(nasium), PettingZoo, scikit-learn, NumPy, pandas, Matplotlib, Git, Spark, Hadoop, Neo4j, QuantLib