Sebastian Lee
Lead AI Engineer at deepset
Landsberg am Lech, Bavaria, Germany
Summary
🤩
Rockstar🎓
Top SchoolSebastian Lee is a Lead AI Engineer with eight years of industry experience and over a decade of combined research and applied work, now driving Haystack’s architecture and multi-agent capabilities at deepset. He builds production-grade RAG pipelines, developer tooling, and observability for generative AI, leading a team to deliver scalable agent frameworks used by thousands of developers. His contributions to major open-source projects like Haystack and Hugging Face Transformers include model optimizations, QA features, and infrastructure improvements such as crawler, training, and similarity-scaling enhancements. Trained as a theoretical chemist (Ph.D. from Caltech), he brings a rare blend of rigorous scientific problem solving and practical ML engineering—evident in prior wins like a Pytorch hackathon project and performance-driven recommender work. He combines hands-on backend and DevOps skills with technical leadership in enterprise deployments, often surfacing subtle model and tooling fixes that improve reliability at scale. Based in Bavaria, he focuses on turning advanced research into production systems that are debuggable, observable, and maintainable.
8 years of coding experience
4 years of employment as a software developer
California Institute of TechnologyBachelor of Science, Chemistry/Biochemistry, GPA: 3.9, Bachelor of Science, Chemistry/Biochemistry, GPA: 3.9 at University of California, Santa Barbara
English, German
Github Skills (18)
pytorch10
language-model10
selenium10
python10
machine-learning10
selenium-ide10
deep-learning10
question-answering10
transformer10
nlp10
architecture9
model-optimization9
modeling9
architectures9
ag8
Programming languages (6)
Github contributions (5)
deepset-ai/haystack
Jul 2022 - 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:
Back-end & DevOps Engineer
Back-end & DevOps EngineerContributions:820 reviews, 148 commits, 285 PRs in 6 months
Contributions summary:Sebastian focused on enhancing the crawler functionality and dependencies, as evidenced by the updates to the minimum Selenium version, gRPCio pin, and the addition of the requests requirement. They also contributed to improving the training process within the Haystack framework by refining the DebertaV2 architecture check within the Trainer.train. The user further addressed issues related to the FARMReader by updating the evaluation to be consistent, implementing the "no answer" logic and updating related documentation. Additionally, they implemented features such as early stopping and a scaling/thresholding mechanism for similarity scores.
pythonbertfinetuningproduction-readyhuggingface-transformers
huggingface/transformers
Feb 2023 - Oct 2023
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:15 reviews, 5 PRs, 31 comments in 7 months
Contributions summary:Sebastian primarily focused on debugging and enhancing the performance of machine learning models within the Hugging Face Transformers library. Their commits include fixes for overflow issues during training of mDeberta models, and the addition of new question-answering models (T5 and MT5), demonstrating their expertise in model architecture and optimization. The contributions involve code modifications across different model implementations. They also added model_parallel=False.
pythonbertspeech-recognitionstate-of-the-artflax
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.