Łukasz Borchmann

AI Researcher

Poznań, Wielkopolskie, Poland20 yrs 7 mos experience

Highly Stable

Key Highlights

Expert in NLP with over 10 years of experience.
Pioneered state-of-the-art models in document understanding.
Recipient of Best Paper Award at SemEval 2020.

Stackforce AI infers this person is a SaaS expert specializing in Natural Language Processing and Machine Learning.

Contact

lukasz@borchmann.ai LinkedIn

Skills

Core Skills

Natural Language ProcessingMachine Learning

Other Skills

Deep learningPythonPyTorchLinguisticsComputational Linguistics

About

ML researcher with 10+ years of experience in NLP and 15+ years in SWE. Today, I am primarily focused on LLMs, LVLMs, and document understanding, which includes developing novel architectures and inference methods. My 2022 PhD thesis focused on NN architectures shifting paradigm toward what is now called GenAI.

Experience

20 yrs 7 mos

Total Experience

4 yrs 1 mo

Average Tenure

3 yrs 8 mos

Current Experience

Snowflake

Staff Research Scientist

Sep 2022 – Present · 3 yrs 8 mos

Trained state-of-the-art models, including Arctic-Text2SQL-R1 (RLVR for SQL generation), Arctic (480B dense-MoE hybrid), and Arctic-TILT (parameter-efficient vision-language model for document understanding). Researched LLM evaluation and developed novel benchmarks, including the MADQA and DUDE. Proposed novel decoding strategies for generating code and structured outputs. Established and drove research collaborations with academic and industrial partners.

Natural Language ProcessingMachine LearningDeep learningPythonPyTorch

Applica

2 roles

Senior Research Scientist

Jan 2022 – Sep 2022 · 8 mos

Deep Learning Team Leader

Apr 2018 – Jan 2022 · 3 yrs 9 mos

Introduced TILT, a state-of-the-art Text-Image-Layout Transformer model that won ICDAR's InfographicsVQA shared task. Trained models winning SemEval-2020 Task 11 and PolEval 2018 competitions. Pioneered few-shot semantic retrieval, introducing methods to extract complex legal clauses using language models. Created DUE, an end-to-end benchmark and dataset for multipage Document VQA evaluation. Received Best Paper Award at SemEval 2020.

Natural Language ProcessingMachine LearningDeep learning

Adam mickiewicz university

Senior Machine Learning Developer

Dec 2021 – Feb 2024 · 2 yrs 2 mos

Proposed pre-RAG retrieval augmented language models for identifying fake news, hoaxes, and disinformation.

Natural Language ProcessingMachine Learning

Press-service monitoring mediów

Language Modeling Specialist

Jul 2016 – Dec 2017 · 1 yr 5 mos · Poznan, Greater Poland District, Polska

Trained entity and aspect-based sentiment analysis models for commercial media monitoring.

Natural Language ProcessingMachine Learning

Various companies

Career since 2005

Jun 2005 – Dec 2017 · 12 yrs 6 mos

Transitioned from a self-taught software developer to a natural language processing expert. Since 2014, I have contributed to various commercial and scientific ML projects across academia and industry, including Adam Mickiewicz University, Poznan Supercomputing and Networking Center, Warsaw University, and Lionbridge.