LLMs Aren’t Just for Chat: Grammatical Error Correction at Scale

Join us on November 25 to learn how Grammarly (now part of the Superhuman suite) has reimagined its Grammatical Error Correction pipeline in the era of LLMs—evolving data generation, evaluation, and inference to power production systems at scale.
✅ Registration: https://gram.ly/3LTEIVi

🚀 Bridging Research and Production: How We Build Scalable GEC Systems

Join Kostia Omelianchuk, Applied Research Scientist, and Lukas Beisteiner, ML Engineer, as they unpack the full scope of Grammatical Error Correction (GEC) from task framing, evaluation, and training to inference optimization and serving high-performance production systems at Grammarly.

🔬 They will discuss:

• The modern GEC recipe: The shift from primarily using human-annotated corpora to semi-synthetic data generation approaches, as well as its impact on model training, evaluation practices, and overall task definition
• LLM-as-a-judge techniques: Developing trustworthy, scalable evaluation systems that minimize manual effort with less human intervention, including tips for prompt engineering, handling positional bias, and creating task-specific judges
• Making it fast and affordable: Serving a 10x larger model than our legacy system, and how Speculative Decoding rewrites the rules of model deployment

🔈 Kostia Omelianchuk is an Applied Research Scientist and Area Tech Lead at Grammarly. He leads applied NLP research in the writing assistance domain, drawing on over a decade of experience and a strong academic background with widely cited publications in Grammatical Error Correction. His current work focuses on multilingual modeling, large language model evaluation, and synthetic data generation.

🔈 Lukas Beisteiner is an ML engineer with 10 years of experience, including 3 at Grammarly. He has worked on a variety of features, including AI detection and the models that power Grammarly’s underlines, focusing on making ML systems production-ready, scalable, and optimized for millions of users worldwide.

💥 This event is geared toward Applied Researchers and ML engineers who are exploring how to build and evaluate LLM-powered systems in practice, as well as developing, deploying, or scaling LLM applications in production.

Agenda:
✨ 18:30–19:00: Registration and networking
✨ 19:00–20:00: Talk
✨ 20:00–21:00: Mingle with our team

✅ Where: In-person, Superhuman Berlin hub
✅ When: Tuesday, November 25
✅ Language: English
✅ Registration: https://gram.ly/3LTEIVi
The event is free. Registration is mandatory. Due to a limited number of seats, the invites will be sent to a limited number of interested guests on a first registered, first invited basis. Please check your inbox for a confirmation email about your attendance.