Adam Lerer
Member Of Technical Staff at Anthropic
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolAdam Lerer is a seasoned machine learning systems engineer with 11 years of experience building high-performance, research-grade infrastructure for large-scale ML and graph problems. He has driven core back-end work across influential open-source projects such as Torch7, cuTorch, PyTorch-BigGraph and OpenNMT—improving memory management, CUDA device APIs, and production-ready embedding pipelines used at scale. Adam has held engineering and research roles at Meta, DeepMind, OpenAI and Anthropic, blending systems engineering, game-theoretic multi-agent research, and production ML deployment. His background includes developing software for the Anton supercomputer and being a founding PyTorch developer, reflecting deep competence in numerical computing, performance optimization, and scalable model training. Notably, his contributions often target low-level reliability and debugging (better error reporting, GC triggers, non-contiguous tensor fixes), a detail that quietly elevates developer productivity and system robustness.
11 years of coding experience
14 years of employment as a software developer
BS, M. Eng., Computer Science, Physics, BS, M. Eng., Computer Science, Physics at Massachusetts Institute of Technology
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (38)
debugging10
pytorch10
debug10
machine-translation10
gru10
neural-machine-translation10
operation10
python10
back-end-development10
tensorrt10
memory-management10
machine-learning10
rnn-model10
c1110
n10
Programming languages (9)
Github contributions (5)
facebookresearch/PyTorch-BigGraph
Jul 2019 - Jan 2022
Generate embeddings from large-scale graph-structured data.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:15 reviews, 55 commits, 11 PRs in 2 years 6 months
Contributions summary:Adam contributed significantly to the `torchbiggraph` repository, primarily focusing on enhancements and additions to the core functionality. They implemented new features for the `TensorList` module, including `clone`, `apply`, `combine`, and mathematical operations like addition and subtraction. Furthermore, the user added support for different edge list types beyond TSV, which involved developing a flexible interface and integrating Parquet support. They also worked on performance optimizations like allocating shared memory at the beginning and made several bug fixes to improve the overall project's robustness.
structuredgnnembeddingsgraph-neural-networkmachine-learning
pytorch/examples
Jan 2017 - Jan 2020
A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc.
Role in this project:Backend Developer
Backend DeveloperContributions:3 PRs, 21 comments in 3 years
Contributions summary:Adam's commits focus on the implementation of a word language model within the PyTorch examples repository. Their initial commit introduces the core structure of the model using RNNs, LSTMs, and GRUs. Subsequent commits modify the model by adding multi-layer capabilities and refining code to align with PyTorch's expected behavior. Finally the user refactors to use the standard torch RNN library and adds a generate.py script.
pytorchvisiondeep-learningreinforcement-learningreinforcement
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.