Master Thesis Student/ Data Science Intern

Thesis title - A comparison of time series models for revenue and product demand forecasting in smart fridges. / Ein Vergleich von Zeitreihenmodellen für Umsatz- und Produktnachfrageprognosen bei intelligenten Kühlschränken.

• Developed an end-to-end Revenue and Product demand forecasting system with a novel ‘storage-oriented’ architecture, thus bringing down the average response time from 8 seconds to 3.5 seconds.
• Introduced efficient data migration strategies (MongoDB, BigQuery), cutting monthly bills by 42%.
• Introduced exogenous variables and dynamic model parametrization components, bringing down the MAPE from 63% to 27%.
• Implemented a ‘data-drift based’ model retraining strategy, eliminating the human in the loop.
• Implemented loggers, test cases and memory profilers for better system diagnostics and monitoring.
• Efficient query optimization and caching for better payload handling.
• Configured CI/CD pipeline – Docker Image build, Test, Security Check, Cloud Run, Deploy, Migrate.
• Setting up API service and configuring Nvidia Jetson Nano for the Google’s ‘Mediapipe’ library for people counting, age and gender detection service. Created a real time dashboard for the same using Dash.
• Achievements - Top Star Performer Award.

Artificial Intelligence Intern

The digitalisation unit of the Krones group, implementing state-of-the-art digital filling and packaging line.

• Formulated strategies for model retraining of the environment model of the company's benchmark Reinforcement Learning based product – ContiloopAI.
• Updated the previous data pipeline to feature engineer raw factory data coming from an end device(ReadyKit) via Share2Act. This helped us to save on-site manual data acquisition and field trips.
• Developed strategies for model retraining via different kind of performance based and time based triggers.
• Documented training articles on reinforcement learning and model retraining and monitoring for the AI team.

Data Scientist/ Data Analyst

SCHMP, Department of Food, Civil Supplies and Consumer Protection (Run by AS Foundation).

• Developed end-to-end machine learning prototypes and scaled them to run in a production environment.
• Created a SVM based classifier system with 92% accuracy to assign ration cards to different consumer based on their income. This reduced the request processing time for a consumer by 60%
• Derived actionable insights from massive data obtained from 52 districts of the state to facilitate better policy making decisions.
• Analyzed old information architectures and contributed to the design and development of new ones.
• Created easy-to-understand visual dashboards projecting growth and giving insights to facilitate the proper execution of government schemes.
• Created a chatbot using IBM Watson for quick redressal of urban consumer grievances.

Bachelor of Engineering - Information Technology

Grade: First class distinction.

Master of Science - Artificial Intelligence and Data Science

Grade: 1.8