Nihar Shah

Experience

• 

  Neewee

  Sept 2020 - Aug 2022

  Data Scientist

  • Constructed and launched a novel time-wrapping algorithm for splitting production processes in a continuous environment; achieved 99% accuracy (+/- 2 sec error) in time segmentation, enabling more precise analysis of production data.• Collaborated with cross-functional teams including BA, engineers, UI developers, and product managers in an agile project to gather user requirements and translate them into actionable data insights; reported customer satisfaction at 93%.• Customized and operationalized a real-time sensor data pattern matching system that identified anomalies and production inefficiencies. This resulted in a 4% increase in the production of good batches.• Utilized historical data and machine learning techniques to generate optimal production rulesets and environmental conditions for each process; reduced production losses by 7%; client saved over USD 1 Million per annum. Show less

• 

  Northeastern University

  Jan 2023 - Jun 2023

  Team Lead

  • Delivered a hands-on workshop on " Power BI for Effective Data Storytelling" to 140 students, enhancing their data communication skills.

• 

  Radical AI

  May 2023 - Dec 2023

  Machine Learning Engineer

  • Developed and deployed a fine-tuned LLaMA 2 based chatbot to reduce data transfer lag time between Vertex AI and chatbot interface by 21%, resulting in a more responsive user experience.• Leveraged prompt engineering to fine-tune a Vertex AI chatbot on personalized data, achieving 98% contextually accurate responses. This significantly improved user satisfaction with the chatbot's ability to address their inquiries.• Spearheaded the integration of ReX, a personalized adaptive learning chatbot, into RadicalX's platform, scaling the daily active user count from 150 to 52,500. Show less

• 

  The 5 Gyres Institute

  Jun 2023 - Sept 2023

  Data Analyst | XN Project

  • Spearheaded the creation and rollout of a real-time data pipeline to feed a comprehensive dashboard. This enabled stakeholders to gain insights on global plastic pollution trends 12.5 times faster than previously possible. • Performed thorough data cleaning and pre-processing to enhance data quality and accuracy for analysis by 17%. This involved identifying and correcting inconsistencies in data sets collected from various sources.