Kazi Sabrina Sonnet

Network automation: Automation of traffic node integration (domain user and profile management, asset management, PIER CR and task development, PIER task scheduling, report generation and audit) into several different network subsystems in the core network service management platform (CNSMP), which increased deployment efficiency, reduced human errors and reduced time to market by 80 percent.

Device return and write-off prediction model: Built machine learning (ML) model that can help reduce inventory costs, enhance free cash flow by continuously predicting high propensity return cohorts, minimize drive-up programs/offers, predict consumer propensity to default payments, and drive write-off reduction programs. With the ML model, the constructive approach to consumer recognition and mitigation lowered returns by 10 percent to 20 percent, which means that the projected savings will be $13 million - $26 million per year.

Development of KPIs for launching 5G Standalone Mode:  Developed an algorithm to analyze the trend of call model for future dimensioning and addressing any immediate issues that predicts the traffic handled by Service-
Communication Proxy (SCP) functions at the peak time of the day. Automation and visualization were done by deploying the script to Grafana dashboard.
 

Development of an anomaly detection algorithm from statistical analysis of Diameter Routing Agent (DRA) Nodes: Developed an anomaly detection algorithm by statistically analyzing the trends of requests and responses from diameter protocol based servers and clients (4G Technology) using Python, Pandas, Influxdb and Grafana dashboards to monitor daily traffic of DRA nodes for determining if any change of configuration is
required for any of the nine DRA sites.
 

Software development for Service-Communication Proxy (SCP) audit: Developed several multi-threaded software for SCP audit (5G Technology) to match the network routing information of twelve different SCP sites and generate alarm if any discrepancy is detected.
 

Development of a web-frame to investigate real-time virtual service data: Developed a web app using Django and Python to scrutinize the queries from REST APIs in real-time.​

Predicting Movement Disorder Symptoms Using Machine Learning Algorithms: Developed a binary and multi-class approach implementing machine learning algorithms to evaluation of tremor severity based on data remotely recorded by a mobile- or tablet-based drawing application. This work has been used to differentiate between healthy subjects, essential tremor patients receiving deep brain stimulation
treatment, and those same patients with treatment disabled. This classification algorithm was highly effective, with overall accuracy of 97.04%.
 

Android Application Development: Developed a self-administered and noninvasive telemonitoring application using Android Studio in Java
programming language and Java Scripts that enables the Essential tremor patients to collect data at home, performs real time statistical analysis and transmit the result over the internet to a dedicated server. This application has been used to collect data from subjects that enables the researchers to develop various tests which cannot be easily performed with paper and pencil.

Windows Application Development: Designed a Graphical User Interface (GUI) using Windows Presentation Foundation (WPF) in C# to monitor and analyze real-time brain signal from Medtronic's Nexus-D3 and Activa PC+S products. This investigational research tool has been used in the closed loop deep brain stimulation (CLDBS) study, to establish a connection between the implanted neurostimulator (Activa PC+S®, Medtronic INC) in Essential tremor patient and an external computer​

Designed and developed analog and digital circuitry including schematic creation, component selection, PCB layout for signal processing electronics and amplifiers for radiation detection systems.