Alexandre Marques
Machine Learning Engineer at Neural Magic
Somerville, Massachusetts, United States
Summary
🤩
RockstarAlexandre Marques is a Machine Learning Engineer with a decade of experience specializing in numerical models and computational algorithms, now based in Somerville, MA. He focuses on making ML models practical and performant in production, with hands-on work improving sparsity-aware inference in the widely used deepsparse runtime. Alexandre’s contributions include fixing transformer prediction and question-answering pipelines and refining evaluation tooling (e.g., MNLI matched/mismatched splits), showing attention to both correctness and benchmarking. Comfortable bridging research-grade numerical methods and production inference, he brings a pragmatic mindset to optimizing CPU inference workloads and debugging complex model pipelines.
10 years of coding experience
Github Skills (10)
transformers10
machine-learning10
nlp10
inference10
python10
onnx9
pre-trained-model8
pruning6
quantization5
computer-vision5
Programming languages (2)
Github contributions (5)
neuralmagic/deepsparse
May 2022 - Sep 2022
Sparsity-aware deep learning inference runtime for CPUs
Role in this project:
ML Engineer
ML EngineerContributions:49 reviews, 8 commits, 19 PRs in 4 months
Contributions summary:Alexandre primarily focused on improving the deep learning inference runtime, addressing bugs and enhancing the functionality of the `deepsparse` library. Their work included fixing prediction access issues in `transformers/eval_downstream.py` and the question-answering pipeline in `src/deepsparse/transformers/pipelines/question_answering.py`. Furthermore, the user contributed to the evaluation scripts by splitting the MNLI dataset evaluation into matched and mismatched groups and adding dataset kwargs. This indicates a focus on performance optimization and model evaluation.
llm-inferenceruntimetensorflowsparsificationmachinelearning
neuralmagic/lm-evaluation-harness
Jul 2024 - Mar 2025
A framework for few-shot evaluation of language models.
Contributions:89 pushes, 7 branches in 8 months
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.