Machine Learning Data Science

Kiran Neupane

**** ******* ****, ********, ** ***01 — 573-***-****— ************@*******.***

LinkedIn: linkedin.com/in/kiranneupane/ — github.com/kngbq

Qualifications Profile

Skilled and results-driven Data Science and Analytics, and Software Development professional seeking a chal-

lenging position to apply advanced knowledge in data analysis, machine learning, and statistical modeling.

With a solid foundation in data science methodologies, programming proficiency, and a passion for extracting

actionable insights from complex data, I am eager to contribute to a forward-thinking organization where I can

utilize my expertise for data-driven decision-making and solve complex problems.

Key Strengths - Detail-oriented, innovative, passionate about research and development advances in data sci-

ence and machine learning. Continually completing required tasks, training, and troubleshooting, Quick learner

Technologies:

Environments: Linux, Windows, MacOS

Tools: Python, R, SQL, NoSQL, Jupyter Notebook, Hadoop, Spark, TensorFlow, Scikit-learn,

Keras, PyTorch, Kubeflow, Tableau, Power BI, React, Nodejs, Flask, javascript, HTML,

CSS, Kubernetes, Docker, Kubeflow, Bash, Github, JIRA, Angular, Springboot

Cloud-Platforms: Amazon Web Services, Google Cloud Platform, Nautilus and GENI (Research Clouds)

Skills: Statistical modeling, AI applications, Data Engineering, Performance Evaluation, Inter-

personal Skills

Professional Experience

Graduate Research Assistant, University of Missouri August 2021 - Present

Involved in various research and development works as a GRA in CERI Center assuming different roles such as

that of a Data Analyst for projects relating to cybersecurity, health informatics and data journalism, and as a

software developer for development and maintenance of web applications. Research works focused on applying

machine learning tools and techniques across varied disciplines. Implementation of a neural network for the

classification of malicious and benign network traffics, application of machine learning classifiers in the Smart

Scheduling application for No-Shows for health appointments, and development of algorithm to process emails

and classify based on their newsworthiness. Usage of ETL techniques to create several projects that necessitated

the utilization of data engineering skills.

• Implemented a custom sequential neural network model that outperformed available open-source LLM models

for its scale including BERT variations, aiding editorial decision support for a local newsroom. Developed

an AI model for sorting email pitch based on its newsworthiness leveraging text extraction using TF-IDF

vectorizer, keyword extraction, topic modeling using LDA/NMF factorization and Sentiment Analysis.

• Automated crime reporting systems improving the efficiency of news dissemination and timely reporting

leveraging ETL techniques on police blotters.

• Initialized cost-cutting tasks to reduce the cost of implementation around automation and machine learning.

Application of machine learning algorithm using a multi-label perceptron as a main neural network to classify

types of attacks with an ensemble of different machine learning models, malicious network traffic flows from

benign ones based on cost, impact, and calculating risk assessment.

• Development and maintenance of LMS systems viz., Mizzou Cloud DevOps and Mizzou Cyber Range for

teaching Cloud DevOps concepts and Cybersecurity using tools such as Angular and Springboot.

• Developed a model for a Smart Scheduling healthcare application where we can schedule appointments of

patients based on available records. Applied hyperparameter tuning to improve the model for an increased

accuracy metric by 20% applying SMOTE techniques to normalize the under-sampled No-show appointment

class. Developed a machine learning pipeline and created documentation for an LMS at the University.

• Aided in grant writing tasks for probable projects involving cybersecurity, and bioinformatics that involved

hypothesis testing and statistical modeling.

• Created visualization for the evaluation sets of models, dashboards for computational demands and network

traffic to communicate the results based on their performance and helping decision-making.

• Leveraged parallel computing to process large image datasets using Kubernetes in Nautilus Cloud and ap-

plying Convolutional Neural Networks for pattern recognition.

Graduate Student Coordinator, Research Experience in Undergraduates May 2023 - July 2023

Led the REU program and Summers@Mizzou “Hacker Tracker” initiatives, overseeing the education of

undergraduate researchers and high school students. Offered guidance and mentorship to undergraduates,

addressing their research concerns and providing support. Mentored participants in a two-month program

and supervised research graduates, enabling their work to be published in prestigious conferences. Under my

supervision, undergraduates successfully authored five papers applied to top-tier conferences.

Academic Tutor, Mizzou Athletics August 2021 - May 2022

Tutored college athletes at Mizzou Athletics in Statistics and Mathematics, providing them with personalized

academic support. To enhance their learning experience, created tailored lesson plans and developed course

materials that catered to their specific needs. Additionally, fostered a collaborative relationship with professors

to ensure that the tutoring sessions were in sync with the course curriculum, ultimately helping the athletes

excel in their academic pursuits.

Education

University of Missouri - Columbia Master of Science in Data Science and Analytics, 2023

High Performance Computing

Southeast Missouri State University Bachelor of Science, Business Administration 2020

Research and Development Projects

Associated Press (AP) Local News AI Project Feb 2023 - Aug 2023

Developed cutting-edge AI-based applications tailored for two newsrooms, with a primary focus on enhancing

editorial decision support and automation. Built a neural network, setting a new industry standard by enabling

precise and innovative decision-making for a local television news organization. Code repository accessible from

Github: https://github.com/associatedpress/local-ai-brainerd-dispatch

• Engineered a pioneering custom-trained neural network to aid editorial decision-making for a local news

organization.

• Played a critical role in crafting advanced AI solutions, demonstrating expertise in data science, machine

learning, and collaborative teamwork.

• Contributed significantly to the successful implementation of AI technologies, enabling newsrooms to adapt

to innovative automation solutions effectively.

• Collaborated with AP on developing AI-based applications for two newsrooms, focusing on enhancing editorial

decision support and automation.

CICADA January 2022

Research in active defense reference architecture that leverages decoy systems to deceive attackers based on

the observability of attack with optimized strategy selection based on cost and risk of varying decoy systems

provisioned in Enterprise IoT network. Attack detection using Binary Neural Networks (NN) for classifying

benign and attack traffic. Multilayered NN for further classifying attacks into specific categories including some

of the modern attacks such as Ransomware, Zero-Click.

• Conducted research on an active defense reference architecture, focusing on leveraging decoy systems to

deceive attackers in an Enterprise IoT network.

• Optimized strategy selection for decoy systems based on the observability of attacks, taking into account cost

and risk factors.

• Use of intelligent agents for optimization of attack risk and deployment costs for varying decoy environments

Mizzou Cloud and DevOps August 2021

Created an intelligent scheduling system for healthcare leveraging Mini Kubeflow (ML toolkit), allowing patient

appointments based on existing records. Enhanced model accuracy by 20% through meticulous hyperparame-

ter tuning and utilized SMOTE techniques to balance under-represented No-show appointments. Established

a comprehensive machine learning pipeline and documented a Learning Management System (LMS) for the

University, accessible from: https://www.mizzouclouddevops.net

• Developed a learning module implementing an SVM model that can for health patient appointments.

• Drafted write-ups and reading materials for the use case application in Healthcare.

• Developed interface for the modules.

Crop Loss Indemnity Jan 2022 - Aug 2022

Applied advanced data analysis techniques using R to assess crop loss patterns and trends due to drought,

focusing on states heavily affected in the USA. Played a crucial role in analyzing and visualizing agricultural

data, leading to actionable insights and policy recommendations.

• Developed choropleth maps illustrating crop loss in high-impact states over a 7-year period, pinpointing

regions needing immediate attention and mitigation strategies.

• Conducted in-depth analysis of crop loss trends nationwide, highlighting significant peaks in 2012 and ad-

dressing the positive impact of the Environmental Quality Incentives Program between 2014 and 2018.

• Investigated the correlation between drought-related crop loss and irrigation indemnities, emphasizing the

need for a balanced distribution of indemnities among states facing severe agricultural challenges.

• Collaborated with agricultural stakeholders to advocate for equitable distribution of irrigation indemnities,

promoting fair policies that address the specific needs of high-risk states like Texas, Kansas, Illinois, Iowa,

North Dakota, and South Dakota.

Western Blot Analysis Current

Implementing a Convolutional Neural Network (CNN) for Western Blot Analysis, focusing on analyzing protein

expression levels from images received. Leading the project from data preprocessing to model development and

evaluation, resulting in accurate predictions and valuable insights for biological research.

• Preprocessed raw image data, including noise reduction, image augmentation, and normalization, ensuring

high-quality input for the model.

• Developing a custom CNN architecture tailored for Western Blot analysis, optimizing layers and hyperpa-

rameters to enhance prediction accuracy.

• Collaborating with biologists and researchers to interpret the model’s predictions, providing valuable insights

into protein behavior.

AKDOT Current

• Studying improvements on the efficiency of a port in Alaska involving simulations of optimal routing, rerout-

ing, incident report and response, developing alert applications for timely updates.

Certifications

Kubernetes Administrator, Skillsoft, 2023.

Machine Learning, Propensity Score and Segmentation Modeling, Skillsoft, 2023

Responsible Conduct for Research, CITI Program, 2022

Relevant Links

LinkedIn: https://www.linkedin.com/in/kiranneupane/

Github: https://github.com/kngbq

References

Professor Prasad Calyam,

Director of Mizzou CERI, Greg L. Gilliom Professor of Cyber Security

University of Missouri - Columbia

Mr. Ernest L. Kung,

Project Manager - Associated Press

Available upon request.

Publications

1. K. Neupane et al., ”Automation of News Content Curation and Storytelling for Local Newsrooms.” Ac-

cepted in IEEE DIKW, 2023.

2. R. L. Neupane, T. Zobrist, K. Neupane et al., “CICADA: Cloud-based Intelligent Classification and Active

Defense Approach for IoT Security.” Accepted and presented in IEEE INFOCOM GI Symposium, 2023.

3. K. Neupane et al., ”Experiences with a Virtual Reality System for Immersive Decision Making and Learn-

ing”. Accepted in IEEE AIPR, 2023.

4. K. Neupane et al., “Intelligent Active Cyber Defense in Cloud and Edge-based Systems: A Survey”. Under

Revision in ACM Computing Surveys.

Contact this candidate