• Developed an ML Engine for predicting network outages for Nokia's telecom clients across NA, APAC, and EMEA with 78% accuracy at specific locations.• Integrated three live status and alarm monitoring software for WLAN devices across 550+ Nokia offices into a centralized dashboard, enhancing 65% operational efficiency.• Contributed to 10+ cloud and automation initiatives in collaboration with Nokia's Dallas, Helsinki, Espoo, Shanghai, and Bangalore offices.

Collaborator: National Research Council Canada (NRC)• Collaborated with cross-functional, cross-territory teams from the National Research Council Canada, Hannover Centre of Germany, the University of Ottawa, and Max-Planck Club on a project to develop an alternative to fiber optics for transmitting high-speed data to Canadian northern territories (Yukon, Northwest, Nunavut) economically.• Simulated the metamaterial device with existing software, scanned over many design parameters for data collection, and pre-processed, stored, and organized the collected data.• Trained deep neural networks on the data to create surrogate models that replaced the simulation software.• Developed an AI-based optimization algorithm to solve the inverse problem, which is to specify thethe metamaterial-based device that gives the desired optical response.• Shared insights effectively with interactive visualizations, using Power BI dashboards to help the team identify trends, gain insights from experiments, and make informed decisions.

• Created speaker recognition, audio dictation, and environmental classification systems using MFCC, RNN, and NLP with Google Translate API and deployed them on hearing aid devices, manufactured by this Start-up, using a data pipeline through android apps.• Analyzed consumer data collected with Meta Business Suite to optimize marketing campaigns of WeHear on Facebook & Instagram, resulting in a 150% higher conversion rate.

• Developed a novel spectroscopical device, 'WeSense', for adulteration detection in oil, milk, and medicines that uses multi-class classification models on the dataset, prepared from scratch using spectroscopy and achieved 98% accuracy in the classification and integrated this system with a mobile app.• Generated 18 basic features using the properties of the electromagnetic spectrum, such as reflection rate, for a new non-destructive testing method based on Machine Learning classification models to replace costly and time-consuming High-Performance Liquid Chromatography, a destructive adulteration testing method in analytical chemistry. • Resulted in a 95% accurate identification of adulteration in the given sample in just 90 seconds.

• Worked on a series of research projects in the domain of blockchain and autonomous vehicles by collaborating with international authors and universities, namely, the University of Kentucky and Durban University.• Designed viable solutions by utilizing Blockchain, AI, and its subdomains such as distributed systems, cloud computing, big data, image processing, and cost-optimized development to put forward our solutions and drafted 3 research papers for international journals.• Proposed a hybrid model of LSTM and GRU to forecast the accurate prices of Litecoin and Zcash in four different window sizes by mapping the interdependency between coins and their parent currencies, such as Bitcoin, in order to contemplate market sentiments.• Developed an 86% accurate Deep Learning-based modified neural network to detect any potential bugs, faults, or malfunctioning in the software of an autonomous vehicle and outperformed state-of-the-art techniques.

Participated in creating several ML and Data Science solutions for private and government sector clients, such as a data analytics project for the ‘C-Plan’ application of the UP police department.