Peter Lofgren
Member Of Technical Staff at Anthropic
Burlingame, California, United States
Summary
👤
Senior🎓
Top SchoolPeter Lofgren is a Member of Technical Staff at Anthropic with 12 years of experience building large-scale ML systems and applied LLM safeguards. He previously led high-impact ML work at Stripe, cutting costs by over $20M/year with fraud-blocking models and driving projects that reduced false positives and modernized Radar and merchant fraud systems. His background spans research and production: a Stanford PhD on Personalized PageRank with a 70x speedup on massive graphs, to engineering a real-time graph service and allocator at Teapot that informed open-source graph work. Comfortable across languages and stacks, he has shipped systems in Scala, Rust, C++, Python, PyTorch and TensorFlow and contributed low-level JVM graph improvements to notable projects like Cassovary. Known for translating business problems into robust end-to-end ML solutions, he combines rigorous algorithmic thinking with pragmatic productionization and monitoring. Based in Burlingame, CA, he brings both deep research chops and hands-on engineering to safety-focused LLM development.
12 years of coding experience
11 years of employment as a software developer
University of Minnesota Twin CitiesMaster of Science (MS), Computer Science, Master of Science (MS), Computer Science at Stanford University
Stackoverflow
29reputation
4kreached
1answer
0questions
Github Skills (9)
javas10
java10
graph-theory10
scala10
data-structure9
data-structures9
algorithms8
algorithm6
permutation6
Programming languages (3)
Github contributions (5)
twitter/cassovary
Jan 2015 - Aug 2015
Cassovary is a simple big graph processing library for the JVM
Role in this project:
Backend Developer
Backend DeveloperContributions:19 commits, 11 PRs, 68 comments in 7 months
Contributions summary:Peter primarily focused on modifying and enhancing the `SynchronizedDynamicGraph` class, a core component within the Cassovary library. Their contributions involved addressing issues related to node ID handling (specifically, supporting ID 0 and the `DELETED_MARKER`), refactoring the code for improved efficiency, and correcting minor bugs in the `toString` method. They also implemented a new `ArrayBasedDynamicDirectedGraph` utilizing FastUtil for improved performance and added supporting features to the `GraphReader` to include the new graph type.
graph-processingjvmprocessing-librarygraphjava
plofgren/cassovary
Jan 2015 - Jan 2016
Contributions:67 commits, 14 PRs, 66 pushes in 11 months
graph-processingjvmprocessing-librarygraphjava
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.