Dan Stancl
Senior AI Platform Engineer
Prague, Prague, Czechia
Summary
🤩
Rockstar🎓
Top SchoolDan Stancl is a Senior AI Platform Engineer with 7 years of experience building cloud-first, scalable AI applications and platforms, currently shaping AI infrastructure at Rossum in Prague. He combines product-focused engineering—ranging from pretraining T-LLMs to streamlining model deployments and web services—with a strong open-source pedigree as a core maintainer of TorchMetrics and significant contributions to Hugging Face Transformers (including porting LongT5 and adding head-mask support across encoder–decoder models). His work spans PyTorch, TensorFlow, and JAX, and he’s helped harden PyTorch Lightning through mypy-driven type fixes, showing attention to both research-grade models and production robustness. With an MSc in Computational Statistics & Machine Learning (Distinction) from UCL and earlier experience in semantic search and knowledge-graph research, he bridges academic rigor and product delivery. A detail that often surprises collaborators: beyond model code, he contributes lower-level metric implementations (ROUGE, BERTScore, ChrF++, TER) that materially improve NLP evaluation at scale.
7 years of coding experience
5 years of employment as a software developer
Bachelor's degree, Economics and Finance, with Honours, Bachelor's degree, Economics and Finance, with Honours at Charles University in PragueUniversity College London
Czech, English, German, Slovak
Github Skills (18)
pytorch10
language-model10
python10
mypy10
machine-learning10
pytorch-lightning10
bart10
encoder-decoder10
typehinting10
deeplearning-ai10
deep-learning10
natural-language-processing10
transformer10
type-checking10
nlp10
Programming languages (6)
Github contributions (5)
Lightning-AI/torchmetrics
Aug 2021 - Jan 2023
Machine learning metrics for distributed, scalable PyTorch applications.
Role in this project:
ML Engineer
ML EngineerContributions:539 reviews, 91 commits, 87 PRs in 1 year 5 months
Contributions summary:Dan primarily contributed to the implementation of machine learning metrics within the `torchmetrics` library, focusing on natural language processing (NLP) tasks. They added new implementations of ROUGE and BERTScore metrics, including core functionality like calculating Rouge-N, RougeL, and RougeLSum scores, and integrating BERTScore functionality. They also improved existing metrics, such as by vectorizing parts of the `MeanAveragePrecision` metric, and added support for new metrics like `ChrF++` and `TER`.
machine-learningpytorchpythondata-sciencedeep-learning
huggingface/transformers
Jan 2021 - Aug 2022
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
Role in this project:ML Engineer
ML EngineerContributions:32 reviews, 32 commits, 38 PRs in 1 year 7 months
Contributions summary:Dan made significant contributions to the `transformers` repository, specifically focusing on implementing and testing new features for the BART model. Their work included adding support for `head_mask` and `decoder_head_mask` to BART-based models like BART, MBart, Blenderbot, Marian, and Pegasus. The user also addressed issues related to head masking in encoder-decoder models. In addition, the user was involved in incorporating the `decoder_head_mask` argument into T5 models.
audioinferencemachine-learning-modelsmultimodaltransformers
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.