NLP Engineer

Role

• Implemented Information Extraction algorithms which efficiently run on 114k Electronic Health Records data of patients.
• Developed algorithms with +90% F1 Score to extract, disambiguate, normalize and relate medications and their attributes like dosage, route, frequency, duration, etc.
• Integrated weak supervision techniques, like Skweak and Snorkel, to proliferate training data for SiteofMets extraction; improved the results on test data by 7.5% F1 Score.
• Streamlined annotation with active learning in SpaCy, cutting manual annotation effort by 30%.
• Optimized a sentiment analysis pipeline with AWS Comprehend, Textract, Lambda and S3, reducing processing time by 80%.
• Revamped model agnostic metrics calculation tool and deployed it company-wide.