Michal Grochmal
About Michal Grochmal
Michal Grochmal is a Senior Data/ML/Software Engineer at BenevolentAI and a Visiting Lecturer at City, University of London, with extensive experience in data pipelines and NLP.
Title
Michal Grochmal holds the position of Senior Data/ML/Software Engineer at BenevolentAI and serves as a Visiting Lecturer at City, University of London. He operates in a hybrid work setup from London, England, United Kingdom.
Career Overview
Michal Grochmal has an extensive career in data and software engineering, with experience at notable institutions. He currently works at BenevolentAI and City, University of London. His previous roles include Assistant Vice President Programmer at Bank of America Merrill Lynch, Data Warehouse Architect at Met Traders Ltd, System Programmer at Fidelity National Information Services, and C Developer at Metavante.
Education and Expertise
Michal Grochmal has a strong educational background in computer science and physics. He achieved a Doctor of Philosophy (PhD) in Computer Science from Birkbeck, University of London. He also holds a Master's Degree (MSc) in Intelligent Technologies from the same institution. Additionally, he pursued doctoral studies in the Security of Distributed Systems at Queen Mary University of London and has academic experience from Jagiellonian University and Federal University of Parana in physics-related fields.
Specializations in Data and NLP
Michal specializes in building data pipelines for fetching, filtering, curating, and feeding text, particularly for scientific papers. He manages the unique challenges associated with human-curated data in NLP pipelines, focusing on building uniqueness, primary/foreign keys, and consistency. He designs and writes RDBMS-like operations into distributed systems and emphasizes eventual consistency in data pipelines for models like Named Entity Recognition (NER), Relation Extraction (RE), and Large Language Models (LLMs).
Open Source and Community Involvement
Michal Grochmal actively participates in the GNU/Linux community, contributing to projects and discussions. He is currently working on writing an easy-to-read transformer layer on top of autograd/autodiff to demystify the training phase of sequence learning for less mathematical students. He advocates for using Vim, describing it as 'zen' and ideal for usage with pedals.