Sunil Kumar - DevOps Engineer
Sunil Kumar - DevOps Engineer
Sunil Kumar
SUMM ARY
Detail-oriented Data Science and AI enthusiast with over 6 years of experience in data analysis and
modelling. Proficient in machine learning and deep learning tools such as Python, TensorFlow, and
scikit-learn. Familiar with designing and implementing solutions in Generative AI, LLMs, and NLP.
Committed to using data-driven insights to support decision-making. Seeking opportunities in Data
Science and AI development, with a focus on leveraging analytics to contribute to innovative
business solutions.
SKILLS
Mentor and guide junior researchers, data scientists, and technical teams on AI/ML best practices and
project implementation.
Collaborate with cross-functional teams, including domain experts, researchers, and software developers, to
deliver end-to-end AI solutions.
Big Data & Simulation
Conduct big data analysis using Monte Carlo simulations for data correction and variance reduction in
large datasets.
Implement transport modeling techniques for predictive analytics in detector response and performance
evaluation.
Technical Content Development
Review, generate, and manage technical content related to AI/ML, data modeling, and scientific research
for academic and industry purposes.
Healthcare & Medical Imaging
Apply computer vision techniques for tumor detection and segmentation in medical images (MRI, PET-CT) u
(sing deep learning frameworks like UNET and Mask R-CNN.
Ensure data quality and enhance diagnostic imaging through advanced visualization and processing
techniques.
COMPANY DETAI LS
Working as Research Associate in Panjab University & CERN (Jul 2024 to Present)
Domain Expert in Soul AI Oct 2023 to Apr 2024
Senior Research Fellow in Punjab University & CERN (Mar 2020 to Feb 2023)
Junior Research Fellow in Punjab University & CERN (Mar 2018 to Feb 2023)
Nuclear Medicine Technologist in KCHRC (Aug 2010 to Jul 2013)
PROJECTS-11
LangChain for LLM Application Development (Coursera)
Developed LLM applications using the LangChain framework for personal assistants and specialized
chatbots.
Implemented agents, chained calls, and memory management to enhance LLM functionality.
LangChain: Chat With Your Data (Coursera)
Built a chatbot that generates responses based on the content of provided documents, utilizing Retrieval-
Augmented Generation (RAG).
Developed expertise in LLMs for retrieving and utilizing contextual documents from external datasets.
Legal Text Processing with LLMs
Designed and optimized LLM architectures for processing complex legal documents.
Developed preprocessing pipelines for family law PDFs, ensuring clean data for model input.
Fine-tuned models to improve classification and Named Entity Recognition (NER) accuracy in legal texts.
Predictive Modeling for Cryptocurrency Price Movements
Developed machine learning models to predict cryptocurrency price trends.
Optimized hyperparameters and engineered features to enhance trading strategy recommendations.
Classification Model for Purchase Prediction
Created a classification model to predict purchase interest based on customer interaction data.
Improved engagement strategies for the sales team, leading to optimized conversion rates.
Variance Reduction for Cost Control in AWS
Conducted predictive modeling using Monte Carlo simulation with a weighted bias technique to predict
scatter kernels.
Achieved a 54.65% reduction in simulation time and a 6.62% decrease in data variance, lowering AWS
simulation costs.
Completed as part of a Marie Curie Early Stage Researcher project at Durham University.
Pituitary Tumor Identification
Implemented deep learning algorithms (CNN, UNET, Mask R-CNN, YOLO) for identifying pituitary tumors in
MRI and PET-CT images.
Enhanced model performance using transfer learning and data augmentation, with comparative analysis
of ResNet and VGG backbones.
Computation Cluster Monitoring
Installed and maintained ROCKS 7.0 (CentOS 7.4) on HP clusters for computational research.
Monitored performance and ensured system reliability for large-scale data processing tasks.
Transport Modeling and Big Data Generation
Last Updated: October 3,
Developed predictive models for detector response using statistical and systematic uncertainty analysis.
Analyzed large datasets using C++/ROOT framework and simulated detector behavior through big data
generation techniques.
Finite Element Analysis (FEA)
Performed 3D finite element analysis using ANSYS Mechanical APDL for detector models.
Evaluated symmetry models to reduce the size and cost of finite element simulations.
Data Quality Assurance (CERN Project)
Focused on data analysis and quality assurance as a Visiting Scientist at CERN.
Created visual reports and conducted performance evaluations using data visualization tools.
Image Analysis and Resolution
Developed a PET detection model to enhance image resolution through Monte Carlo simulations.
Applied machine learning for event reconstruction and quantum entanglement analysis.
Q-PET: Quantum Entanglement-Based Positron Emission Tomography
Developed models for quantum entanglement-based PET imaging.
Published research findings in Springer Proceedings of Physics.
Technologies/Skills used: C++, Python, Data Visualization, Data Modeling, MC Simulation, Data Generation
Leading big data analysis for background correction estimation within the CMS collaboration as L3
Simulation Convener.
Working in data quality monitoring using data visualization framework.
SOUL AI
Role: Domain Expert (Oct 2023 to Apr 2024)
Technologies/Skills used: C++, Python, Data Analysis, CNN, ANSYS, Big-data, ML/DL, Data Visualisation
Pituitary Tumor Identification
Implemented deep learning algorithms for identifying pituitary tumors in MRI and PET-CT scan images,
using CNN architectures such as UNET, Mask R-CNN, and YOLO.
Enhanced model performance through transfer learning and data augmentation, with evaluations
conducted on Kaggle datasets.
Conducted comparative analysis of ResNet and VGG backbones for optimal tumor identification
accuracy.
Computation Cluster Monitoring
Installed and maintained ROCKS 7.0 (CentOS 7.4) on the HP cluster.
Transport Modelling and Big Data Generation
Predictive modelling for detector response using transport models, incorporating both statistical and
systematic uncertainties.
Analyzed large datasets with the C++/ROOT framework, conducting comparative analysis between
real-world and simulated results.
Generated big data using statistical probability distribution functions to simulate detector behaviour
and response.
Finite Element Analysis [FEA]
Analyzed the field solution for each mesh node using the FEA method with ANSYS Mechanical APDL for a
3-D detector model
Last Updated: October 3,
KCHRC
EDUCATION
PhD in High Energy Physics (2018)
Panjab University & CERN
CERTIFICATIONS
Deep Learning & Medical Image Analysis
Completed: June 2023 – August 2023
Machine Learning & Artificial Intelligence
Institution: IIT Roorkee, India
Completed: January 2019