Bogdan Kostić
Machine Learning Engineer at deepset, makers of Haystack
Berlin, Germany
Summary
🤩
Rockstar🎓
Top SchoolBogdan Kostić is a Machine Learning Engineer based in Berlin with 7 years of applied ML experience, specializing in NLP and building production-ready LLM applications. At deepset he has moved from intern to machine learning engineer, contributing significant code and tutorials to flagship open-source projects like Haystack and FARM—helping users implement retrieval-augmented generation, QA pipelines, and BERT-style pretraining. His background in computational linguistics and data science (Univ. of Potsdam, BHT, and a UC Berkeley visit) underpins a strong focus on language data processing, evaluation metrics, and token-level sequence handling. Bogdan blends research and engineering, having supported linguistic fieldwork and taught syntax/morphology while also shipping practical notebooks and pipeline integrations for real-world developer adoption. Colleagues describe him as a pragmatic bridge between academic rigor and product-focused ML engineering, with a knack for turning complex language models into accessible tutorials and reproducible workflows.
7 years of coding experience
6 years of employment as a software developer
Visiting Student, Visiting Student at University of California, BerkeleyMaster of Science - MS Data Science, Master of Science - MS Data Science at Berlin University of Applied Sciences Berlin (BHT)Erasmus+ Computer Science & Language and Multimodal Interaction, Erasmus+ Computer Science & Language and Multimodal Interaction at Università di TrentoBachelor of Science - BS Computational Linguistics, Bachelor of Science - BS Computational Linguistics at University of Potsdam
Italian, French, Spanish, English, German
Github Skills (17)
pytorch10
language-model10
python10
lm10
large-language-models10
llm10
retrieval-augmented-generation10
question-answering10
bert10
jupyter-notebook10
transfer-learning10
nlp10
pre-trained-model9
data-handling9
deep-learning8
Programming languages (6)
Github contributions (5)
deepset-ai/haystack
Mar 2020 - Jan 2023
AI orchestration framework to build customizable, production-ready LLM applications. Connect components (models, vector DBs, file converters) to pipelines or agents that can interact with your data. With advanced retrieval methods, it's best suited for building RAG, question answering, semantic search or conversational agent chatbots.
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:7 releases, 447 reviews, 108 commits in 2 years 10 months
Contributions summary:Bogdan primarily contributed to the development of tutorial notebooks demonstrating the use of the Haystack framework for building LLM applications. Their work included adding and integrating Jupyter and Colab notebooks for tutorials, which involved the addition of code snippets. The contributions focused on the implementation of pipelines and demonstrating their functionality.
pythonbertfinetuningproduction-readyhuggingface-transformers
deepset-ai/FARM
Apr 2020 - Apr 2022
:house_with_garden: Fast & easy transfer learning for NLP. Harvesting language models for the industry. Focus on Question Answering.
Role in this project:ML Engineer
ML EngineerContributions:6 reviews, 54 commits, 18 PRs in 2 years
Contributions summary:Bogdan's contributions focused on enhancing the data processing and model training components of the FARM library, specifically for language model adaptation and pre-training. This included implementing token-level sequence handling, integrating BERT-style pre-training, and addressing related bug fixes. Further work involved adding new evaluation metrics and refining the data silo structure for more streamlined evaluation workflows, emphasizing a focus on question answering tasks.
bertxlnet-pytorchnatural-language-processingtransformersharvesting
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.