Lukas Blecher
Research Engineer at Meta
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolLukas Blecher is a research engineer based in Paris with seven years of experience building ML systems for document and mathematical content understanding. Currently at Meta after contributing to Magnit @ Meta AI, he specializes in math OCR and neural optical understanding, combining hands-on model work with backend data processing. His open-source contributions include core improvements to Facebook Research’s Nougat and significant work on LaTeX-OCR (pix2tex), where he implemented robust math extraction, LaTeX generation, and image conversion pipelines. Comfortable across Python tooling, transformers, and image-processing libraries, he focuses on making research-grade models reliable and production-ready. Colleagues describe him as analytically driven—bringing mathematical insight into practical engineering solutions for complex document parsing challenges.
7 years of coding experience
1 year of employment as a software developer
Ruprecht-Karls-Universität Heidelberg
Github Skills (15)
transformers10
regular-expression10
pytorch10
machine-learning10
latex10
deep-learning10
python10
image-processing10
document-processing10
natural-language-processing9
model-optimization9
ocra9
deeplearning-ai9
ocr9
computer-vision9
Programming languages (4)
Github contributions (5)
lukas-blecher/LaTeX-OCR
Dec 2020 - Sep 2022
pix2tex: Using a ViT to convert images of equations into LaTeX code.
Role in this project:
Back-end Developer & Data Scientist
Back-end Developer & Data ScientistContributions:6 releases, 11 reviews, 251 commits in 1 year 9 months
Contributions summary:Lukas's primary focus was on implementing and improving math extraction and LaTeX code generation functionalities. They developed scripts to extract mathematical expressions from documents, including handling various LaTeX environments such as equations, aligns, and inline math, all using regular expressions in Python. The user also worked on generating images from LaTeX code, and the code changes demonstrate a strong understanding of LaTeX and image processing libraries for converting LaTeX formulas into PNGs.
pythonlatexim2latexlatex-ocrmath-ocr
facebookresearch/nougat
Aug 2023 - Feb 2025
Implementation of Nougat Neural Optical Understanding for Academic Documents
Role in this project:ML Engineer
ML EngineerContributions:3 reviews, 20 PRs, 44 pushes in 1 year 6 months
Contributions summary:Lukas primarily contributed to the Nougat project by implementing and refining core functionalities, including fixing package issues, correcting licensing, and adding support for newer versions of the transformer libraries. They also introduced features such as markdown compatibility and failure detection, indicating a focus on improving the usability and robustness of the document processing pipeline. The user's involvement extended to optimizing the model and its integration within the prediction and training workflows.
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.