Summary
Devang Kulshreshtha is an Applied Scientist II at Amazon AWS Agentic AI in New York with 11 years in industry and 5+ years focused on machine learning research and engineering. He leads practical defenses and safety work for large language models, including engineering the Bedrock Guardrails Prompt Attack solution that outperformed competitors by ~35%, and optimizes multilingual long-context inference for production guardrails. His background spans ASR, NLP, agentic AI safety, and lifelong learning—contributing core science to launches like HealthScribe and Lex while authoring papers at EMNLP, INTERSPEECH and IJCAI plus six patents. He has also built scalable ASR personalization for catalogs exceeding 500K items and developed lifelong training methods that reduce the need for full retraining. Trained at McGill/Mila under Prof. Siva Reddy and experienced in edtech question generation at Korbit.AI, he blends rigorous research with production impact. Notably, he combines deep academic publication experience with hands-on product wins that measurably improve safety and speech systems at scale.
11 years of coding experience
5 years of employment as a software developer
95.6 %, 95.6 % at Shanti Niketan Public School
St. Clares Senior Secondary school
Bachelor's degree, Computer Science and Engineering, 9.11, Bachelor's degree, Computer Science and Engineering, 9.11 at Indian Institute of Technology (Banaras Hindu University), Varanasi
Master's degree, Computer Science, 4/4, Master's degree, Computer Science, 4/4 at McGill University